unguaiat. it... B aim x v I: 11118.! .2, L33...- r5 (,1... am»: lira-n a.» .1! I. 5.11:2; f ,. u) it n. 0‘ r 1 .1. act... ’3an . n' 9%..» .1 5...... .- .. s. . .I A .. hz.u.h.2.{.itlv - l)‘:-§>’-I3§!‘I 34h? .L ., PuT:.§J..4.i§¢mr s w . vL61. 9|... fimfii PS c. .1 av. < L. flm -a...nn.uF€:tu. 3. (M. 1:13.14: .3 . J) 4.. :uryh 1.. . .5 .9. (It in )I' ($1.3 I31}... ‘1’: .Afi. "Va... ,5, \ I 35. v b. r. MICHIGANS SATET ! RABIES 3! !!!!!!!!!!!!!!!!!!!!!!!!!!!l W!!! 2364 (f "I” ‘7’ 5‘ !! This is to certify that the dissertation entitled Automation and Control of the Microwave Processing of Composites. presented by Valerie Omega Adegbite has been accepted towards fulfillment of the requirements for PhD . degree in Chemical Engineering WKW? Major professor Date Wfir MS U is an Affirmative Action/Equal Opportunity Institution 0-12771 LIBRARY Michigan State University PLACE II RETURN BOX to remove this checkout tron: your record. TO AVOID F INES return on or baton duo duo. DATE DUE DATE DUE DATE DUE %'= 9% M80 IoAnNfirmntlvo MM“ Oppmilylmtltulon AUTOMATION AND CONTROL OF THE MICROWAVE PROCESSING OF COMPOSITE MATERIALS By Valerie Omega Adegbite A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirement for the degree of DOCTOR OF PHILOSOPHY Department of Chemical Engineering 1995 ABSTRACT AUTOMATION AND CONTROL OF THE MICROWAVE PROCESSING OF COMPOSITE MATERIALS By Valerie Omega Adegbite Microwave processing of composites in a sin gle-mode resonant cavity using fixed frequency technology, was automated to advance the current state of technology by bridging the gap between device and process. Automatic hardware and software for controlling the non-linear and complex electromagnetic interactions during composite curing were developed. Efficient coupling, uniform and controlled heating were the control objectives which were achieved through mode switching, mode tuning, and power control. For mode tuning a non-traditional control methodology using a 2-dimensional simplex minimization method was used, and shown to be more efficient than manual univariate methods. In the development of the uniform heating controller theoretical empty cavity solutions were used to develop empirical correlations to characterize the loaded cavity. This was necessary to overcome the computationally intensive calculations required for solving loaded cavity equations for control purposes. Using these empirical correlations, mode switching for achieving uniform or controlled heating was constructed and shown to be highly effective. Power control was based upon traditional proprotional-intergal-derivative(PID) control methodology and shown to be more proficient than the previous on/off control. The developed controllers were integrated into an overall closed-loop feedback control system that was implemented in a control program called LabView. Another control system was also built and demonstrated by interfacing with a knowledge-based system planner. The control systems were implemented in a newly designed automated cavity with novel mechanized drives and axial and radial mounted microwave coupling probes. Composites curing showed that the ease and flexibility in operating the automated microwave process was comparable to automated thermal processes, although it contains extensive electronics and involves complex control tasks. Compared with the manually operated system, sample temperature gradients were reduced by 60%, mode switching times Were reduced by 67%, mode tuning reproducibility was significantly improved and data acquisition was enhanced for processing and for diagnostics. This marks a notable step in the advancement of the current state of technology; from that of a manual lab-scale device to a fully automated “First generation prototype” process. The application potential of this technology has been greatly enhanced by this development. To my parents, Victor and Jeannette Adegbite, my siblings, Curtis Adegbite,! Alvina Adegbite-Obuobi, Gwendolyn Adegbite-Cooper, for the amaranthine and unconditional love and support. ACKNOWLEDGMENTS I wish to extend my sincere gratitude to Dr. Martin Hawley, my advisor, for his guidance and support throughout the course of this work. His editorial and extensive technical insights on the development of the dissertation has enhanced it to a fine and comprehensive document. Credit is due to Dr. Asmussen, whose wealth of knowledge in electromagnetics provided the fundamental understanding in this research. There is also credit due to Dr. Jon Sticklen who provided the computing insight that steered the direction of this work. I wish to also extend my thanks to Dr. Radcliffe for sharing his knowledge in process control and for introducing the software tool that was used in this work. Thanks is also due to Ron Fritz, Dale Wesson, Larry Fellows, Jianghua Wei, Jim McDowell, David Decker, Yunchang Qiu, Michael Muczynski, and Beajaye Bedell for their various assistance through the course of this work. This research was funded by NSF I/UCRC Polymer Processing Center at Michigan State University. The computing hardware and software were provided by ISL (Information System Laboratories) at Michigan State University TABLE OF CONTENTS LIST OF TABLES ............................................................................. xiii LIST OF FIGURES ............................................................................ xiv Chapter 1: INTRODUCTION ................................................................... 1 Chapter 2: REVIEW OF PERTINENT LITERATURE 2.1 Introduction ............................................................................. 14 2.2 Control Philosophy .................................................................... 16 2.2.1 Conventional I Traditional Control .............................................. 17 2.2.2 Intelligent Control ................................................................. 17 2.3 Composite Materials ................................................................... 20 2.3.1 Background ........................................................................ 20 2.3.2 Fabrication Methods .............................................................. 20 2.4 Control in Composite Processing -Autoclave ....................................... 21 2.4.1 Composites Process Modeling ................................................. 21 2.4.2 Conventional and Conceptual Control Methods ............................... 21 2.4.3 Intelligent Control Methods ...................................................... 21 2.5 Control in Composite Processing -Microwaves ................................... 26 2.5.1 Process Modeling ................................................................. 26 2.5.2 Control in Microwave Processing .............................................. 27 2.6 Data Acquisition ........................................................................ 27 2.6.1 Autoclave ....................................... 27 2.6.2 Microwave ............................................................................ 27 2.7 Summary ................................................................................ 28 Chapter 3: ELECTROMAGNETIC THEORY AND MICROWAVE PROCESSING 3.1 Introduction ............................................................................. 29 3.2 Electromagnetic Theory ................................................................ 29 3.2.1 Maxwell's Equations ............................................................. 29 3.2.2 Maxwell’s Equations in a Source Free Homogeneous Region .............. 31 3.3 Electromagnetic Fields in a Cylindrical Region ..................................... 32 3.3.1 Empty Cavity Solutions .......................................................... 32 3.3.1.1 Modes ......................................................................... 35 3.3.1.1.] Transverse Electric Modes - TE Modes ............................. 36 3.3. 1.1.2 Transverse Magnetic Modes-TM-modes ............................ 37 3.3.1.2 Modes Designation .......................................................... 38 3.3.1.3 Electric Field Pattern ........................................................ 39 3.3.1.4 Cut-Off Frequency .......................................................... 39 3.3.1.5 Cavity Quality Factor ....................................................... 46 3.3.2 Loaded Cavity Solutions ......................................................... 48 3.3.2.1 Small- Low loss Samples .................................................. 48 3.3.2.2 Homogeneous and IsotrOpic Samples .................................... 49 3.3.2.3 Simplified Composite Material ............................................. 49 3.4 Fundamentals of Microwave Heating ................................................ 50 3.4.1 Microwave Power Absorption and Perrnittivity ............................... 50 3.4.2 Penetration Depth ................................................................. 51 3.4.3 Applicators ......................................................................... 53 3.4.3.1 Multimode and Waveguide ................................................. 53 3.4.3.2 Single—Mode Cavity ......................................................... 54 3.4.4 Mode Tuning ...................................................................... 55 3.5 Summary ................................................................................ 56 Chapter 4: EMPTY AND LOADED CAVITY CHARACTERIZATION S 4.1 Introduction ............................................................................. 57 4.2 Empty Cavity Characterization ........................................................ 58 4.2.1 Experiments ....................................................................... 58 4.2.2 Results and Discussions - Empty cavity characterizations ................... 58 4.3 Loaded Cavity Characterization ....................................................... 61 4.3.1 Experiments ....................................................................... 61 4.3.1.1 Frequency Shift Measurement ............................................. 63 4.3.1.2 Electric Field Measurement ................................................. 64 4.3.2 Results and Discussions - Loaded cavity characterizations .................. 67 4.3.3 Loaded cavity mode estimation .................................................. 89 4.3.3.1 Results and Discussions .................................................... 91 4.4 Summary ................................................................................ 94 Chapter 5: AUTOMATION CONTROL SOFTWARE AND HARDWARE 5.1 Introduction ............................................................................. 96 5.2 Cavity Automation ..................................................................... 99 5.2.1 Cavity Description ................................................................ 99 5.2.2 Automated Cavity and Mechanized Drives ................................... 100 5.2.2.1 Stepper Motor and Driver Hardware .................................... 108 5.2.2.2 Stepper Motor Driver Software .......................................... 110 5.2.3 Cavity Length and Probe Depth Measurement .............................. 112 5.3 Microwave System Automation .................................................... 113 5.3.1.1 Automation of External Circuit .......................................... 113 5.3.1.2 Automation of Microwave Power Source .............................. 117 5.4 Data Acquisition ...................................................................... 118 5.4.1 Data Acquisition Hardware .................................................... 118 5.4.2 Data Acquisition Software Prnoram .......................................... 118 5.5 Process Control Software ........................................................... 121 5.5.1 Efficient Coupling- Mode Tuning ............................................. 121 5.5.1.1 Approach ................................................................... 121 5.5.1.2 Simplex Method ........................................................... 123 5.5.1.3 Simplex Logic ............................................................. 124 5.5.2 Mode Selection and Uniform Heating ........................................ 130 5.5.2.1 Approach ................................................................... 130 5.5.2.2 Mode Selection Logic ..................................................... 131 5.5.3 Temperature Control ........................................................... 134 5.5.3.1 Approach ................................................................... 134 5.5.3.2 PID Method ................................................................ 134 5.6 LabView Curing Process Control Software Program ........................... 136 5.6.1 Front Panel ...................................................................... 136 5.6.2 Program Implementation ...................................................... 138 5.7 Diagnostics System Automation .................................................... 140 5.7.1.1 Hardware Automation .................................................... 140 5.7.1.2 Diagnostics system software development ............................. 142 5.8 LabView Software Interface with Knowledge-Based-System Planner ....... 143 Chapter 6:APPLICATION TO CURING 6.1 Introduction ........................................................................... 146 6.2 Curing Experiments .................................................................. 146 6.3 Results and Discussions ............................................................. 148 6.3.1 24-ply Sample ................................................................... 148 6.3.1.1 Mode Selection ............................................................ 148 6.3.1.2 Power Control ............................................................. 149 6.3.1.3 Tuning ...................................................................... 149 6.3.2 48-ply Sample ................................................................... 150 6.3.2.1 Mode Selection ............................................................ 150 6.3.2.2 Tuning ...................................................................... 151 6.3.2.3 Power Control ............................. 151 6.4 Summary and Conclusions .......................................................... 152 Chapter 7: SUMMARY AND CONCLUSIONS 7.1 Introduction ........................................................................... 160 7.2 Automatic Control Software ........................................................ 162 7.2.1 Mode Tuning Software Program .............................................. 162 7.2.2 Mode Selection and Uniform Heating Control Software Program ........ 163 7.2.3 Power Control Software Program ............................................ 164 7.2.4 Data Acquisition Interface and Control Software Platform ................ 165 7.2.5 Data Acquisiton Software program ........................................... 165 7.3 Automation Hardware ............................................................... 166 7.3.1 Cavity and Circuit ............................................................... 166 7.4 Verification of Control System ..................................................... 167 7.5 Diagnostic System Automation ..................................................... 168 7.6 Understanding to Enhance Utilization of Technology ........................... 169 7.6.1 Dual coupling .................................................................... 169 7.6.2 Coupling Probe Effects ........................................................ 170 7.6.3 Loaded Resonant Cavity Length .............................................. 171 7.6.4 Effect of Tooling ................................................................ 171 7.6.5 24-ply versus 48-p1y curing ................................................... 172 7.6.6 Theoretical Field Pattern Plots ................................................. 172 7.6.7 Application of Knowledge-based System to Automation .................. 173 7.7 Global Conclusion ................................................................... 174 Chapter 8: RECOMMENDATIONS AND FUTURE WORK 8.1 Automated System ................................................................... 176 8.2 Technology Advancements .......................................................... 177 8.3 Variable Frequency ............................... 177 8.4 High power source ................................................................... 178 8.5 Dual Coupling ........................................................................ 179 8.6 Scale-up ............................................................................... 180 8.7 Tooling Issues ........................................................................ 180 8.8 Potential Applications ................................................................ 181 Appendix A: AUTOMATION HARDWARE A.1 Stepper Motors ....................................................................... 183 A11 Description ...................................................................... 183 A.1.2 Stepper Motor Driver .......................................................... 183 A13 Wiring Diagram ................................................................. 186 A.2 Data Acquisition Interface ........................................................... 188 Appendix B: DUAL-COUPLING B.1 Introduction ........................................................................... 191 8.2 Experimental .......................................................................... 191 8.3 Results and Discussions ............................................................ 195 B4 Conclusions .......................................................................... 198 Appendix C: KNOWLEDGE-BASED SYSTEM INTERFACE C.l Introduction ........................................................................... 200 C2 Goals of the Planner ................................................................. 200 C3 Interface with KBS—planner ........................................................ 202 Appendix D: ELECTRIC FIELD PATTERNS D.1 Description ......................................................................... 204 D.l.l TM-mode ........................................................................ 204 D.l.2 TE-mode ......................................................................... 204 Appendix E: SOFTWARE DOCUMENTATION E.l Introduction ........................................................................... 205 E.2 Compcure3-Curing program ........................................................ 206 E3 Surfacevi - implementation of tuning program .................................. 206 E4 Replace.vi-replacement of simplex triangle vertices ............................. 206 E5 Vertex.vi -calculation of intial vertex in the simplex triangle .................... 206 E6 CL / PD / Pwr -scaling of cavity length, probe depth and power values ...... 207 E7 Move Read.vi - adjustment of cavity length & probe depth and power sensing207 E.8 Lirnits.vi - tuning limits for a mode ................................................ 207 E9 Contractvi - contraction of simplex triangle in the tuning program ............ 207 E. 10 Pwrcontrl.vi - power controller used in the curing program .................. 208 ' LIST OF CITED WORKS ................................................................... 209 LIST OF TABLES Table 3-1 Penetration depth of various materials ................................................ 53 Table 4- 1 Measured and calculated resonant cavity lengths .................................. 60 Table 4- 2 Results from axial field measurements .............................................. 84 Table 4- 3 Results from axial field measurements .............................................. 84 Table 4- 4 Frequency shift measurement results ................................................ 85 Table 4- 5 Total cavity shift for a loaded cavity for a TM(012) mode ........................ 91 Table 4— 6 Graphite Epoxy Heating Results ..................................................... 93 Table 4- 7 Polyester/ glass heating results ...................................................... 93 Table 6-1 Loaded Cavity resonant length and estimated heating sites ...................... 147 Table 6-2 24-ply curing results ................................................................. 150 Table 6-2 48-ply curing results ................................................................. 152 Table A- l Jumper Setting for Data Acquisiton Board ....................................... 190 Table A- 2 Data Acquisition Board Terminals ................................................ 191 Table B-1 Graphite/ epoxy heating data ....................................................... 196 LIST OF FIGURES Figure 1- 1 Control Loop for the Single-mode Resonant processing System ............... 11 Figure 1- 2 Overall Control Logic for Processing System ..................................... 12 Figure 1- 3 Knowledge-based System Interface Logic ......................................... 13 Figure 2- 1 Feedback control system showing traditional and non-traditional control methods ........................................................................................ 18 Figure 3- 1 Cylindrical coordinate system ....................................................... 33 Figure 3- 2 Elecu'ic field pattern for TE-modes .................................................. 42 Figure 3— 3 Electric field pattern for TM-modes ................................................ 43 Figure 3- 4 Mode chart for TE-modes ............................................................ 44 Figure 3- 5 Mode chart for TM-modes .......................................................... 45 Figure 3— 6 Q.curve calculation using half power point method ............................... 48 Figure 4— 1 Sample Placement ..................................................................... 62 Figure 4- 2 Frequency Shift Measurement ....................................................... 64 Figure 4— 3 Electric field diagnostic holes: a) front view, b) side view, c) top view ........ 65 Figure 4— 4 Axial electric field pattern for empty cavity ........................................ 69 ' Figure 4- 5 Axial electric field pattern for Teflon loaded cavity (a) Elevated (b) Lowered ........................................................................................ 70 Figure 4- 6 Axial electric field pattern for Graphite / epoxy at perpendicular fiber direction (a) Elevated (b) Lowered .......................................................... 71 Figure 4- 7 Axial electric field pattern for Graphite / epoxy at parallel fiber direction (a) Elevated (b) Lowered ......................................................................... 72 Figure 4— 8 Axial electric field pattern for Nylon (a) Elevated (b) Lowered .................. 73 Figure 4- 9 Axial electric field pattern for Polyester/ glass (a) Elevated (b) Lowered ...... 74 Figure 4- 10 Radial electric field pattern for empty cavity ...................................... 75 Figure 4- 11 Radial elecuic field pattern for Teflon (a) Elevated (b) Lowered ............... 75 Figure 4- 12 Radial electric field pattern for Graphite / epoxy at perpendicular fiber direction (a) Elevated (b) Lowered .......................................................... 76 Figure 4—13 Radial electric field pattern for Graphite / epoxy at parallel direction (a) Elevated (b) Lowered ......................................................................... 77 Figure 4- 14 Radial electric field pattern for Nylon (a) Elevated (b) Lowered ............... 78 Figure 4— 15 Axial electric field pattern for Polyester I glass (a) Elevated (b) Lowered. . . .79 Figure 4- 16 Frequency shift measurement for Graphite! epoxy perpendicular fiber direction (a) Elevated (b) Lowered .......................................................... 80 Figure 4- 17 Frequency shift measurement for Graphite / epoxy parallel fiber direction (a) Elevated (b) Lowered ......................................................................... 81 Figure 4— 18 Frequency shift measurement for Nylon (a) Elevated (b) Lowered ........... 82 Figure 4- 19 Frequency shift measurement for Polyester/ glass (a) Elevated (b) Lowered 83 Figure 5- 1 Components of a Typical Single-mode Resonant Cavity. From clock wise, shorting plate and drive, cavity body, base plate with fingerstock, coupling probe 101 Figure 5- 2a Schematic of manually operated single-mode resonant Cavity ............... 102 Figure 5- 2b Picture of manually operated sin gle-mode resonant Cavity ................... 103 Figure 5- 2c Drive of manually operated single-mode resonant Cavity ..................... 104 Figure 5- 3a Schematic of automated sin gle-mode resonant cavity ......................... 105 Figure 5— 3b Picture of automated single-mode resonant cavity ............................ 106 Figure 5- 3c Drive of automated single-mode resonant cavity .............................. 107 Figure 5- 4 Stepper Motor and Driver .......................................................... 109 Figure 5- 5 Program Logic for Stepper Motor Driver ........................................ 111 Figure 5- 6 LabView Program Version of Figure 5-6 ........................................ 111 Figure 5- 7 Linear Motion Potentiometers ..................................................... 113 Figure 5- 8 Schematic of External Circuit ..................................................... 114 Figure 5- 9 Picture of Automated Microwave Processing System .......................... 115 Figure 5- 10 Magnetron Power versus Voltage Calibration for Analog Control .......... 117 Figure 5- 11 Control loop from Chapter 1 ..................................................... 120 Figure 5- 12 Simplex Diagram .................................................................. 127 Figure 5- 13 Simplex Logic Flow Sheet ....................................................... 129 Figure 5- 14 Mode Selection Program Logic .................................................. 133 Figure 5- 15 Program Flow Logic (also in Chapter 1) ....................................... 139 Figure 5- 16 Low Power diagnostic system ................................................... 141 Figure 5- 17 Logic for Interface with planner (Figure 1-2) .................................. 145 Figure 6- 1 Curing temperature profile for 24-ply graphite epoxy ........................... 154 Figure 6 2 Power profile for curing 24-ply graphite epoxy composite .................... 155 Figure 6- 3 Cavity length and probe depth for curing 24—ply graphite epoxy .............. 156 Figure 6- 4 Curing temperature profile fro curing 48-ply graphite epoxy ................. 157 Figure 6- 5 Cavity length and probe depth for curing 48-ply graphite epoxy .............. 158 Figure 6- 6 Power profile for curing 48-ply graphite epoxy ................................. 159 Figure A— 1 Stepper Motor and Drive Component ........................................... 184 Figure A- 2 Stepper Motor Driver TTL Circuit ................................................ 185 Figure A- 3 Wiring Schematic for Stepper Motor Drivers ................................... 187 Figure B- 1 Dual-Couplig Single mode Resonant Cavity ................................... 191 Figure B- 2 Sample placement with thermal paper ............................................ 193 Figure B- 3 Temperature probe placement ..................................................... 194 Figure B- 4 Top probe - graphite epoxy at CL=19.1 cm ..................................... 197 Figure B- 5 Side probe - graphite epoxy at CL=19.7 cm .................................... 197 Figure B- 6 Both porbes graphite epoxy at CL=19.7 cm .................................... 198 Figure C- 1 Global Composite Manufacturing Architecture ................................. 201 CHAPTER 1 INTRODUCTION Microwave heating is a form of electroheating which spans the frequencies of 300MHz to 3OOGHz in the electromagnetic spectrum. The standard microwave frequencies for industrial applications range from 400 MHz to 40 GHz in different parts of the world(Pozar, 1991). In industrial applications in the United States the standard frequencies are 915 MHz and 2450MHz, while other frequencies ranging from 2450MHz to 8000 MHz are used in research applications (Metaxas, 1992). Using the 2.45 GHz frequency, microwave heating of polymer composite materials has been investigated in the waveguide, multimode oven and in the single-mode resonant applicator (Adegbite et a1. 1995; Adegbite et a1. 1993; Adegbite et a1. 1992; Adegbite et a1. 1992; Yunchang et a1. 1995; Yunchang et al. 1995; Shidaker et a1. 1995; Shidaker et al. 1995; Shidaker et al 1995; Wei et al 1991; Lee and Springer, 1984; Lee and Springer 1984; Dhulipala et a1. 1992; McNeil et al. 1992; Wei et a1. 1991; Agrawal and Drzal, 1989; Jow 1988; Wei et a1 1989; Wei et a1 1992; Wei et a1. 1992; Wei et a1. 1992; U. Hottong et a1. 1991; Hawley and Wei, 1991; Wei et a1. 1990; Fellows and Hawley, 1992; Fellows et al. 1993; Fellows et a1. 1994; Mijovic and Wijaya, 1990; Mijovic et a1. 1992; Gourdenne, 1982; Lewis et a1. 1987; J ow et a1. 1987; Jow et a1. 1989; DeLong et a1. 1989; Lewis et a1. 1988; Vogel et a1. 1989; Wilson and Salerno, 1978; Strand, 1980; Gourdene et a1. 1980; Gourdenne and Van, 1981; Karmazsin and Satre, 1985; Jow et a1. 1988;Ju11ien and Valot, 1985; Wei et a1. 1990; Chen and Lee, 1989; Mijovic and Wijaya, 1990). Microwave heating, unlike conventional heating is dependent upon the applicator and sample geometry and properties. Also, unlike conventional heating it involves energy absorption and conversion to heat rather than heat transfer through conduction and 1 2 convection. For non-magnetic materials, microwave power absorption is a function of the microwave power excitation frequency, complex permittivity of the material, and the magnitude of the electric field strength inside the material. The microwave excitation frequency is typically constant and determines the depth of wave penetration into the material. This power absorption relationship is described by Poynting’s theorem as shown in Equation (1-1) (Pozar 1991). The absorbed microwave energy excites the molecules of the sample through dipole rotation and ionic and ohmic conduction which results in the conversion of the microwave energy into heat (Lewis 1992). Thus, in order for a material to be heated by microwaves it must have a complex permittivity which is a description of the dipolar, ionic or ohmic properties. The complex permittivity is described mathematically as a complex quantity. See Equation (1-2). where: P £0 a) C I! § E 0' 1 , P = Swear: 'IE|2 I ' II 2:8 -Je 2 =3 +— soar = microwave power absorbed = perrnitivity of free space = angular frequency = dielectric loss factor = complex permittivity = electric field strength = electrical conductivity (1-1) (1-2) 3 The real part is called the dielectric constant, which is related to the energy stored in a material, and the imaginary part is called the loss factor, which is related to the energy dissipated in the material as heat through the motion of dipoles and charges. The loss factor is a function of material chemical and electrical properties, frequency, temperature and pressure. Materials with a higher loss factor are more easily heated by microwave energy. In general, materials undergo physical and chemical changes during heating . In microwave heating, these sample changes can significantly alter the coupling between the sample and the electromagnetic fields in such a way that process adjustments are required to compensate for them. In microwave processing, the type of applicator defines what processing adjustments must be made. Three different types of microwave applicators are used: the waveguide, the multimode oven and the single-mode resonant applicator. A waveguide is a rectangular or cylindrical hollow pipe which can be used to guide electromagnetic waves. For a ! given waveguide, the type and number of modes that can be excited are fixed. Thus, the waveguide is not conu'ollable to compensate for varying material changes such as the size, shape and especially, changes in complex permittivity. The commercial home microwave oven is what is known as a multimode oven (Asmussen, 1987). In a multimode oven, several electromagnetic modes are randomly _ excited simultaneously for a given applicator volume. Mode stirrers are sometimes used to optimize the excitation of theses modes in the multimode oven (Huack, 1969). The intent is to generate several electromagnetic heating modes such that varying sample parameters during heating are randomly compensated for without adjusting the multimode oven. For a given multimode applicator, the various modes that can be excited may be known, however, the type of modes that are excited at any time are unknown and cannot be controlled. Similar to the waveguide, this restricts the processing 4 capability of the multimode oven to samples such as those that do not vary very much in time and space. A sin gle-mode resonant applicator is designed such that different electromagnetic modes are excited one at a time, by either fixing the frequency and varying the cavity volume, or fixing the cavity volume and varying the frequency. The number of excitable modes is only limited by the length of the cavity or the frequency range of the microwave source. For a given single mode-resonant applicator operating at a fixed microwave frequency, a finite number of electromagnetic modes can be selectively excited, but one at a time, and controlled by the adjustment of the cavity volume. This allows for selective and controlled heating and a potential for application to a wider range of samples of varying chemical and electrical properties and shapes. Using the waveguide, Springer and Lee (Lee and Springer, 1984) processed unidirectional 32-p1y, continuous graphiteepoxy laminates. They reported that only unidirectional composite materials can be heated with microwaves using linearly polarized TEM waves with a polarization angle of 90 degrees. Using a multimode or commercial microwave oven, Lee and Springer (Lee and Springer, 1984) investigated the processing of graphite / epoxy and glass / epoxy laminates. They reported that curing glass epoxy was effective, but the curing of multidirectional graphite epoxy composite was not possible and the curing of unidirectional graphite epoxy depended upon the polarization angle. Chen and Lee(Chen and Lee, 1989) studied the cure of graphite I epoxy and graphite / PEEK(polyether ether ketone) in a single-mode resonant cylindrical cavity using a TE(112) mode at 2.45 GHz. They concluded that the coupling of interactions between microwave energy and composites depended on the fiber orientation and sample geometry in a complex manner. Using a single-mode resonant cylindrical cavity, Vogel (V ogel et. a1. 1989) demonstrated that a 3- inch square, 24-ply graphite epoxy composite can be processed 5 with low input power, and that the heating rate and uniformity were dependent upon the electromagnetic processing modes. Using a single-mode resonant cavity, Wei(Wei, 1989) showed that both unidirectional and cross-ply, thin and thick section graphite epoxy composite materials could be successfully processed using hybrid modes. Also using the single-mode resonant cavity, Fellows (Fellows, 1991) successfully processed polyimide graphite composite panels and planar and complex shaped polyester glass composite materials using a mode switching technique. Using the single-mode resonant cavity, benefits of microwave processing of polymeric composites were reported to range from enhanced mechanical properties, such as in enhanced glass transition temperature of cured epoxy (Wei et a1. 1992; Wei et a1. 1990), enhanced fiber/ matrix interphase properties (Agrawal and Drzal, 1989) faster processing times, and capability to control temperature excursions(J ow, 1989; Wei, 1992). Although these benefits are significant in the composite processing industries, microwave processing of composites is still concentrated in the research arena. Currently in composites processing, the microwave research has been focused in the demonstration of the microwave technology at the lab scale, using specialized toolingCTeflon, ceramic), and through intensive and cumbersome manual operations. In processing, the microwave cavity was Operated as an open-loop system where a seasoned operator was the necessary link to close the control-loop. A typical processing activity 3 included; 1) the continuous tuning of the cavity by the manual rotation of dial knobs to adjust the cavity length and probe depth to minimize the reflected power; 2) the frequent seleCIion of new cavity modes by carefully adjusting the cavity length and probe depth to new locations, while tuning the cavity and manually modulating the input power to achieve uniform heating; 3)temperature control by the automatic on/off control of an electronic switch to direct the microwave to or away from the cavity. Data acquisition was done as a separate activity using both analog and digital ( RS- 232) interface which required complicated program device drivers. Hence, the cavity was operated as an 6 independent device from the external circuit, data acquisition unit and processing results varied from one operator to the other and optimization was difficult at best. In order for the microwave technology to be recognized as a viable technology, it must be demonstrated as a process by addressing processing issues such as specialized tooling, scalability, controllability and automation. Microwave heating in general is an electronic process which typically contains extensive electrical components and control requirements that are unique to the technology. In the single—mode resonant cavity technology, while controllability is one of the attractive attributes it is also a challenge to accomplish for the advancement of the technology. As such, automation and process control which include, feedback systems, data logging, and instrumentation are essential requirements in the advancement of the single-mode resonant cavity as a process. In this work, a control system was designed and built as a process control program was developed and implemented. Two different control software programs were developed; one included complete control logic to meet all of the process control objectives, and the other included only data acquisition, hardware and interface instructions to facilitate an interface with a knowledge-based system planner. The control logic and the overall control loop are shown in Figure l- l and Figure l- 2, and the KBS-interface control logic is shown in Figure 1- 3. Finally, the complete system was demonstrated by curing thin and thick section graphite/epoxy composite materials, and the partial system was demonstrated by interfacing with a Knowledge-based system(KB S) planner to control the processing of epoxy/ graphite composites. In a single-mode resonant applicator, the microwave cavity volume and the coupling probe depth are adjustable variables used to control the electric fields in the cavity. Hence, the automation of this system included the design and fabrication of an applicator with mechanized drives, instrumentation and a closed—100p feedback control system to meet the following control objectives: 7 Efficient coupling (Mode tuning) - Maintenance of cavity resonance in the presence of dynamic sample and process changes by minimizing the reflected power from the applicator. Uniform heating- Control of electromagnetic modes such that energy is coupled optimally to all sample regions. Controlled heating - Regulation of input power such that the sample temperature and reaction exotherrn is controlled and maintained to within 5 °C of the cure setpoint. In general, the significance of this work is realized in the advancement of the single-mode resonant technology by proving it to be a practical process, thus enabling the this technology to be used in the processing of polymer composites and in other applications such as dielectric analysis, plasma and ceramics processing. Specifically, significance would be realized in: 1. The design and implementation of hardware and software for the automation of the single-mode resonant cavity for the convenient and practical operation of it as a process. The development and implementation of a closed-loop feedback control system using traditional and non-traditional control methodologies, to control the electromagnetics inside the cavity during the microwave curing of composites by mode tuning, mode switching and power control to achieve uniform and controlled heating. The use of a 2-dimensional mathematical search technique called simplex as a novel approach for automatically tuning the cavity, by simultaneously manipulating the cavity length and probe depth, to achieve efficient tuning as opposed to the typical univariate or manual methods. 8 4. The development of empirical loaded cavity correlations to understand and control the elecuic fields inside the cavity to achieve uniform heating. 5. The analog regulation of the input power source using proportional- integral-derivative(PTD) controller as opposed to the typical on/off controller (where in the off- position the power was wastefully directed to a dummy load), to control the exothermic temperature excursions to within 5° C. 6. Development and implementation of a control system for interfacing with a knowledge-based(KBS) planner. 7. The design and fabrication of an automated single-mode resonant cavity with novel mechanized drives for cavity length and probe depth adjustments, and instrumentation for automatic position measurements. 8. Automatic data acquisition for fast, reliable and convenient data tracking and maintenance. Dissertation Layout In chapter 2 the concepts of traditional and non-traditional process control methodologies are discussed with emphasis on differences between the two methods and when each method is applicable. A literature review of the current state of automated composite materials processing in an autoclave using these methods are also presented. In general, composite material fabrication is a manual operation which incorporates empirical and fundamental knowledge. In the automation, emphasis is placed on the development of cure cycles dependent upon available process models and automatic sensing devices. In chapter 3, fundamental theoretical background of electromagnetic and microwave processing is discussed. Derivation of field equations and electromagnetic 9 mode classifications, theoretical electric field plots, and their application to a resonant cavity are presented. Empty and loaded cavity solutions for simple cases are also presented. Additionally, the application of these fundamentals to the microwave processing of anisotmpic and'inhomogenous composite materials is also discussed. In chapter 4, experiments and results for empty and loaded cavity characterizations are presented. These experiments include low power diagnostics for empty and loaded cavity modes and the heating of previously cured composite materials for mode characterization. The results in this chapter define the empirical foundation upon which the uniform heating control strategy is deve10ped. In chapter 5, the process control strategies used to meet the different control objectives are presented. The mode tuning controller, was treated as a mathematical problem in which a 2-dimensional simplex search technique was used. This approach is significantly unique because it uses a two dimensional tuning technique rather than a one dimensional technique, by simultaneously adjusting the cavity length and probe depth. Because loaded cavity solutions for non-homogenous and non-isotmpic materials like composites are computationally prohibitive, an empirical method was used to deve10p the uniform heating controller. For the power controller , PID control software was used to manipulate the input power. Each control strategy was deve10ped independently and then integrated to form the overall control system. In chapter 6, results of applications using the developed automated cavity and the developed process control software are presented for the curing of 24-ply and 48-p1y graphite epoxy composite material. During processing, the efficiency of automatic cavity tuning, mode selection for uniform heating, and optimum temperature control are evaluated. 10 Finally, results and conclusions are summarized in chapter 7 and recommendations and future work are discussed in chapter 8. Control program documentation , detailed hardware instrumentation and wiring, dual-coupling results, KBS-interface and other supporting results are discussed in the appendices. 11 . . .- .822 — 830e, 880:0“ _ CZ , Loses— eoeocum . Il— §o2 .2505 - , 5%: Bo: , : 3.32.50 , 88.2 .836 III— mcefi 8oz , 59.3 .335 . : 0:89va Lor— EEEA A Lee—cm 058.85.th _ F3 II.— 882 29% .5680 953.3th , 582 Bo: .822 Samoa 5:2..50 - Swen; .9350 egos—om one: .................... m 3:89va .2 M I I'— GObK—ai— — .333 H . . -uemeom 05.83th L E 2288th ”T. , . . r __ Shams—2 I b.3550 Bison— . - - - 9.69035. Figure l- 1 Control Loop for the Single-mode Resonant processing System 12 K—J Acquire and Scale Daval ® Yes [Power Controller I ! No ' Yes Mode Selection Controller No l . o - @ N Mode Tuning Controller K / Figure l- 2 Overall Control Logic for Processing System 13 _.I Acquire and Scale Data Temperature Power Mode ( KBS Planner] Power Level Mode Power Controller l l I Mode Tuner l Figure l- 3 Knowledge-based System Interface Logic CHAPTER 2 REVIEW OF PERTINENT LITERATURE 2.1 Introduction Microwave processing of composites in a single-mode resonant cavity has been studied as an alternative to thermal processing methods( Fellows, 1992 ; J ow, 1988 ; Lee, 1984;Voge1, 1989; Wei, 1991; Adegbite et. a1. 1995). Results have shown benefits that include enhanced mechanical properties, increased glass transition temperature, increased fiber matrix pr0perties and increased reaction rates (Jow, 1988; Wei, 1991 ). However, the concept of control of microwave curing of composites in a single-mode resonant cavity is novel. The dynamics of the microwave curing process is governed by complex non-linear, time variant interaction between discrete electromagnetic modes and the material inside the cavity. Hence mathematical models that describe the curing dynamics were found to be incomplete and computationally intensive for control purposes. Process monitoring capability of material related pr0perties were also unavailable which further complicated the accurate understanding of the cure dynamics for control purposes. This control problem however, is not unique and fits a class of control t0pics that is an active area of research. Composite curing in an autoclave is an example of a domain that shares similar control problem. Since the primary goal of this work is to develop and implement a control system for the microwave curing process, literature pertinent to control philos0phies and applications in composites processing in an autoclave are reviewed. 14 15 2.2 Control Phi1050phy Control systems can be described as either closed loop or open 100p configurations which correspond to feedback and feedforward, respectively. In feedback systems, the disturbance to the system is not measured but the output is measured and used to regulate the input to the system. In the feedforward structure the disturbance to the system is measured to produce a matching corrective action and the error in the feedforward controller is not checked. Hence, the performance of the controller may deteriorate with time because of parameter drifts(Auslander et a1. 1974). However in the feedback structure there is no need to measure the disturbance since the error of the controller is always checked against its desired value. Hence, a feedback controller is flexible, relatively insensitive to external disturbances, able to function in a changing environment and can even cope with unanticipated disturbances(Auslander 1974). This marks the motivation for the use of the closed loop, feedback control system in this work. Historically, the first conventional feedback control device was the water clock that was invented by the Greek Ktesibios in Alexandria Egypt around 3rd century BC (Mayr 1970). The first mathematical model to describe plant behavior for control purposes was J.C. Maxwell, of Maxwell’s equations in electromagnetics (Antsaklis, 1994 ). In 1868 Maxwell used differential equations to explain instability problems encountered with James Watt’s flyball governor. The governor was introduced in 1769 to regulate the speed of steam engine vehicles. Since the period of the flyball governor, conventional control theory has been advanced with the use of; frequency domain methods and Laplace transforms in the 19305 and 19405; the development of optimal control and state space analysis in the 19505 and 19605; optimal control in 19505 and 19605; and stochastic, robust and adaptive control methods in the 19605 to the concepts of non-traditional control methodologies today. 16 2.2.1 Conventional / Traditional Control The term conventional (or traditional) control is used to refer to the theories and methods to control dynamical systems which are described mathematically by differential or difference equations. In order for the conventional controller to be successfully implemented, the mathematical models must be accurate and simple enough so that they can be solved in real-time. Today, the growing complex structure of chemical processes due to the increasing demand for better energy and materials management, has resulted in demanding control specifications for increasingly complex dynamical systems. 1 This has introduced control problems that cannot be adequately described in a differential of difference equation framework. The control problems mainly pertain to the area of uncertainty due to poor or lack of knowledge or incomplete models to avoid computational complexity. Examples of such systems include discrete event manufacturing and communication system5(Antsaklis, 1994). To address these control problems non-traditional control methodologies known collectively as intelligent control systems were developed. 2.2.2 Intelligent Control Intelligent control systems are non-traditional control methodologies that were developed to address control problems that otherwise cannot be solved by conventional control methods. They are designed to autonomously achieve high level goals, while its components, control goals, plant models and control laws are not completely defined(Antsaklis, 1994). At minimum, an intelligent control system must have the ability to sense the environment and make decisions to control it. Higher levels of intelligence may include the ability to recognize objects and events, to represent knowledge in a model, and to reason about and plan for the future. In advanced forms, intelligence provides the capacity to perceive and understand, to 17 choose wisely and act successfully under large variety of circumstances (Antsaklis, 1994 ). As such, there can be several levels of intelligence control systems. Compared with the traditional control methods, intelligent control systems typically use heuristics and empirical data in form of decision making units rather than predictive models. For many intelligent control systems the controller construction methodology is largely heuristic and based on certain principles from artificial intelligence(Antsaklis, 1994 ). Figure 2-1 shows a control loop for what would be considered to be a traditional and non-traditional control system. The construct of traditional feedback control is preserved while the function of the 18 controllers vary from a predictive one to a search one for the traditional and non- Savoim 5“” * 3333:“ —-—p Controller Process Measured variable Heuristic Empirical Decision Making I 3.. n 11! l 3 PID Process Models Predictive Figure 2- 1 Feedback control system showing traditional and non- traditional control methods traditional approach, respectively. Hence, intelligent control can be described as an enhancement of traditional control methodologies to solve new challenging control problems. Currently, there are several installed intelligent control systems that can be found in the NTST’s (National Institute for Standards and Technology) real-time 19 control system implementations, where most of the systems are for aerospace and military applications(Antsak1is, 1994). Composites materials fabrication is another domain which presents a framework for intelligent control applications. 2.3 Composite Materials 2.3.1 Background Composites are combined materials of two or more components with properties superior to that of the individual constituent materials. The two constituents are typically a fiber and a matrix, where the fiber is the load carrying material and the matrix (polymeric material) holds the fibers in place and provides the bonding between the fibers. Fiber reinforced polymer composites are widely used in aerospace, commercial, military, and other engineering applications. The most significant advantage of composites is the high strength to stiffness per unit weight compared with conventional materials like steel(Strong, 1989; Richardson, 1987 ). 2.3.2 Fabrication Methods The goals for fabricating a composite part is to achieve maximum mechanical properties which requires uniform extent of cure, ultimate part consolidation for uniform resin distribution and minimum void content. Fabrication of composites requires reaction of the matrix system and pressure for the consolidation of the fiber and matrix into a composite part. The matrix can be a therrnosets which are generally liquid resins which undergoe exothermic reaction to achieve a crosslinked network structure. In such systems heat is required to activate the reaction and additional heat is generated as the polymerization reaction proceeds. The heat of reaction can cause the temperature to rise beyond the capacity of the external heating source, whereby in uncontrolled systems runaway reactions can occur and cause thermal degradation of the part The matrix can also be a thermOplastic which are 20 solids which are melted, formed, and then cooled to achieve a solid structure. Thermosets have historically been the principal matrix material for composites although therrnoplastics use is now increasing in many applications(Strong 1989). 2.4 .Control in Composite Processing -Autoc1ave 2.4.1 Composites Process Modeling To overcome this problem a very slow heating rate is typically used which leads to very long cycle times and does not always guarantee uniform heating, especially in thick-section parts. Typically, a predetermined time dependent processing plan is deve10ped off-line and used to control the part temperature and pressure during the cure cycle. The characteristics of the cure cycle depend upon material properties of the resin, fiber, fabrication technique and mold geometry. Hence complete mathematical models that describe the dynamics of the cure cycle are required. The most complete models of autoclave process curing was deve10ped by Loos and Springer in 1983(Springer, l987)which was later improved by (Kardos, 1983;Ha1pin, 1983; Gutowski, 1987; Dave, 1987). The model consist of five sub- models for describing the thermochemical effects, flow consolidation, residual stresses, void formation, strength and modulus. The inputs to this program are mold geometry, material properties, applied temperature and pressure, and vacuum bag pressure as a function of time. The outputs include temperature, pressure, degree of cure, resin viscosity in the part as a function of time and position and the number of compacted prepreg plies as a function time. The strength of this model resides in its ability to predict many aspects of the curing process ranging from heat transfer and chemical reaction during processing, to strength and modulus of the final part. However, because of the intensive 21 computations required these models were used in the simulation of the cure behavior and not for real-time computation for control purposes. Additionally, the limited sensing capabilities of other sample prOperties such as extent of cure and viscosity in the autoclave also complicates the accurate implementation of these models. 2.4.2 Conventional and Conceptual Control Methods A conventional real-time feedback computer control for the autoclave that was deve10ped in 1981 by Applied Polymer Technology Incorporated (Hinrichs, 1984). It was an automated system for controlling the curing of composite structures through an interactive, computerized feed-back system. It included real-time sensors for part surface point temperature and gradient (thermocouples), autoclave ambient temperature(thermocouples), ambient pressure(pressure transducers), and material viscosity (ultrasonic technique). The control decisions were made by comparing the measured values to predetermined optimal values. The pre-determined optimal values were temperature vs. time, pressure vs. time, and viscosity vs. time data which were generated from fundamental cure process models. Up to date this is the only patented automated autoclave system for the curing of composite materials using conventional feed-back control technique. In a theoretical approach (Wu, 1990) used knowledge based systems techniques to develop automated cure cycles for composite manufacturing in an autoclave. In this method process models and experience gained from past Operations were used to provide an initial operation profile and on-line adjustments during the curing. Automation of the autoclave process was simulated using the Loos and Springer models (Springer, 1987) and the process control was simulated by the deve10ped expert system. There are other developed systems which have been verified by controlling an autoclave to cure a part 22 In another approach to autoclave automation, (Trivisano, 1992) deve10ped a method for the control and optimization of the evolution of the processing variables during the fabrication of composite materials. The method allowed for the computation of real-time heat transfer coefficients for each selected tool, the prediction of the temperature changes as a function of the programmed air temperature and the Optimization of the cure cycle to minimize the difference between the actual temperature and recommended values. During real time processing, model predictions were compared with Optimum values to achieve proper control decisions. The significant feature of this architecture is the real—time computation of the tool thermal prOperties which is used Optimize the cure cycle. This may be described as the adaptation of the hardware conditions to optimize the predetermined cure cycles. In the other control and automation approaches, the strong dependence on mathematical models alone was not emphasized while the use of experiential knowledge was studied. These approaches utilized the concepts of knowledge-based systems techniques or intelligent control methods to study the use of experiential knowledge alone or integration of experiential knowledge with mathematical models in the automation and control of the autoclave. 2.4.3 Intelligent Control Methods In 1986 Servais, Lee and Browning(Servais, 1986 ) surveyed the limitations 3 and benefits of the various possible approaches of autoclave control based upon non- fundamental methods. Experimental or trial and error methods were found to be expensive, inefficient, time consuming, inflexible, impractical for optimization, and not practical for on-line control. Thus, suggesting the integration of mathematical 23 methods, and expert systems techniques as the most efficient approach for describing and controlling the curing process in a comprehensive manner. Roberts(Roberts, 1987)pre5ented a comprehensive architecture in which experience, intuition, and mathematical models were integrated in the automation and control of the autoclave. The control architecture included a cure model, a model validation unit, an expert system, and a cure device control unit. Each module in the control architecture was assigned a specific task and was developed independently. The cure model unit was responsible for predicting the Optimum cure profile to be implemented by the expert system. The expert system was the cure Operator which was allowed to adjust the cure cycle based on a set of rules drawn from a model. It also had the capability for determining when there is a need for operator intervention. The device controller was responsible for controlling the process and identifying anomalies in the sensed data that deviated from the predicted cure profiles. It also was responsible for interactively generating process history data files on-line. This unit used the concepts and equipment of the automated control system developed by Applied Polymer Technology(Hinrichs, 1984 ). This is one of the most comprehensive architectures in the automation of the autoclave. Only part of the cure model had been successfully proven, but the total integration of all the modules had not been implemented. One of the unmentioned difficulties that could be experienced is software interface issues, computing power requirements if all of the modules were implemented on one computer, and communication issues if all the modules were implemented on multiple computers. The approach is very robust and implementation will be challenging. In the past ten years some of these non-traditional automation and control concepts have been implemented for composite curing in an autoclave. One is a prototype expert system for controlling the autoclave cure process Lee(lee, 1987). . 24 The control knowledge base contained declarative rules that were derived from an expert about the processing conditions and constraints. This was different from a conventional type approach in which it would require explicitly prescribed temperature and pressure cycles. In another approach a rule based expert system was developed for controlling the autoclave temperature and pressure during cure of fiber reinforced thermosetting composite materials. Rules were established to ensure that the temperature and pressure remained within required limit the residual stresses minimized. The inputs to the expert system were the measured instantaneous autoclave temperature and pressure, the composite midpoint and surface temperatures, composite thickness, and ionic conductivity. These inputs were examined by pre-established rules to control the autoclave settings. The interfaces and rules were incorporated in an algorithm called SECURE and was written in "C" language and installed on a Macintosh computer. The system was verified through model simulations and by controlling an autoclave successfully. In another approach a control strategy was derived from a methodology called a "self directed" approach(LeClair, 1989). In this approach the process was divided into a series of stages and mics were derived from knowledge and intuition of experts to describe each stage. Included in these rules are knowledge about how to change control parameters from the data provided by the sensors, and the conditions required for successful completion of each cure stage. The advantage of this system is that it can be developed without detailed knowledge of all the events that occur in the part. The success of these non-traditional methods is depended upon the quality and quantity of knowledge and the accuracy of data interpretation by the knowledge. These non-traditional automation and control methodologies have classical drawbacks in that they tend to be inefficient, inflexible, and impractical for optimization. This may be credited to the inherently limited size of the knowledge 25 base and the control strategies which tend to be specific and do not allow for a wide variety of processing opportunities. A system in which both mathematical models and knowledge based systems techniques are integrated to control a process was developed by Pardee (Pardee, 1987), at Rockwell Intemational. The system was used for the pyrolysis of carbon- carbon composites. It runs on two Macintosh computers in Allegro Common Lisp. It consist of a control system workstation which provides information about real- time process conditions, and a materials workstation which provides calculated real- time material properties, predicted future properties, estimated current and future index, and an on line expert description and interpretation of the current process state. It has a capability to interpret sensed data and to solve a set of eight coupled differential equations for reaction kinetics information. 2.5 Control in Composite Processing ~Microwaves 2.5.1 Process Modeling The key in microwave process model development is dependent upon the ability to successfully develop a power absorption model to describe the interaction of the electromagnetic fields and the samp1e(see chapter 3). One of the first microwave curing process models was developed by Lee and Springer (Lee and Springer, 1984) using a simplified power absorption model. The output of this model was temperature distribution, the resin viscosity, degree of cure of the resin, resin content of the composite and bleeder cloth during microwave curing. Another process model was also developed using a more rigorous five-parameter estimation power absorption model (Wei, 1991). There are other process modeling developments underway to completely describe the microwave curing process(Sudaram, 1994 ). However, even with the simplified power absorption 26 models, these process models are computationally intensive and applicable to unique processing conditions. The only known complete microwave power absorption model is that developed by IBM(International Business Machines). Unfortunately, published literature is not available about this system at this time. 2.5.2 Control in Microwave Processing The extent of literature found is a clear indication of how widely the intelligent control problem has been studied, especially in the curing of composites in an autoclave. As previously mentioned, the concept of control in the microwave curing of composites is novel. The only known and published automated single- mode resonant cavity was that developed by Alliouat(Alliouat, 1990) and his co- worker in France for sintering ceramic materials. The control system was based upon elements of intelligent control for regulating the input power and for tuning the cavity. 1 A gradient search method was used for tuning the cavity where only sensed information about the cavity length and reflected power were required. Components of this processing include an infrared pyrometer for measuring the surface temperature, detectors for sensing the input, reflected and absorbed power. Controlled parameters were the microwave power supply by analog output, and stepper motors for adjusting the cavity volume. The automatic sensing and control were facilitated by a data acquisition board which resides on an Apple II computer. 2.6 Data Acquisition 2.6.1 Autoclave In the autoclave process, typically sensed data are temperature and pressure of the oven. However, unmeasurable process states such as extent of cure and changes 27 in viscosity are the needed information. Various cure monitoring methods have been studied, which include; ultrasonic sensors to measure viscosity and porosity to describe the process cure states(Trttman, 1988 ); the use of electrodes mounted as roven threads to sense resin flow properties in order to describe process states(Walsh, 1990 ); the use of acoustic ultrasonic technique to measure viscosity and relate it to process sates (Saliba, 1992 ); the use of infrared transmitting Optical fibers to measure the disassociation and formation of chemical bonds (Young, 1988); and - dielectric analysis to measure chemical changes due to molecular motions (Ciriscioli, 1989 ; May, 1983; Day, 1986;Kranbeuhl, 1987). Of all of these methods the dielectric technique has been shown to be most promising for cure state monitoring. At the current state of the technology it can only provide "event" information and um historical information during the cure cycle. (i.e. where viscosity is minimum or when cure is completed). The limited sensing capability combined with the computationally intensive process models provides a framework for studying various control methods. 2.6.2 Microwave Similar to the autoclave process, sensing technology in microwave processing is also limited. There are power meters for sensing the power incident towards the cavity and that reflected from the cavity. The sensed reflected power is used to tune the cavity to resonance, and the sensed input power is used to regulate the amount of power coupled to the cavity. The only direct sample parameter measured is the temperature. This is done using fiber optic therrnometry which is an invasive sensing method(see Appendix B). Unlike the autoclave process, the desired process parameter for control is the microwave curing process is the magnitude of the electric field strength (see chapter 3). Currently, electric field measurement technology is not commercially available and modeling techniques are highly complex and 28 computationally extensive, especially for control purposes. For cure state monitoring, on-line dielectric analysis has been studied at the research level using the sin gle-mode resonant cavity (J ow, 1989). However, this technique is not available at this time for on-line cure state measurement. 2.7 Summary Although several control issues have been widely studied in the autoclave composite curing, control is a very novel concept in the microwave curing process. The control goal set forth in this dissertation is therefore, a significant challenge which will result in a noteworthy technical contribution in the field of control and in microwave processing, especially in the curing of composites. CHAPTER 3 ELECTROMAGNETIC THEORY AND MICROWAVE PROCESSING 3.1 Introduction In this chapter relevant fundamental and theoretical background of electromagnetics and microwave processing are discussed. Derivation of field equations, electromagnetic mode classifications, theoretical electric field plots and their application to a resonant cavity are presented. Empty and loaded cavity solutions for simple cases are also presented along with the application of these fundamentals to the microwave processing of anisotropic and inhomogenous composites. The discussion in this chapter provides the fundamental understanding from which the automation and control software programs were developed. It addresses the concept of modes, mode tuning, cut-off frequency, theoretical elecuic field patterns in the cavity, and empty cavity analysis which are important in the understanding of the loaded cavity. Throughout the derivations, vector quantities will be denoted by a boldface. 3.2 Electromagnetic Theory 3.2.1 Maxwell's Equations Electromagnetic theory at the macroscopic level is embodied in the mathematical equations known as Maxwell's equations (Harrington, 1961). The differential form of the time-varying Maxwell's equations in the general form are shown in Equation (3-1). 29 30 38 V E =-— x at VID=p D=£E (3,1) VxH=-@-+J J=OE at VOB=O B=uH where E is the electric field strength (Volts / meter) D is the elecuic flux density (Coulombs / square meter) H is the magnetic field strength (Amperes/ meter) B is the magnetic flux density (Volts - seconds / square meter) J is the electric current density (Amperes/ square meter) p is the electric charge density (Coulombs / cubic meter) 8 is electric permittivity of the medium (Coulombs/ Volt meter). 0 is electric conductivity of the medium (Coulombs - meter] Volt - seconds) It is the magnetic permeability of the medium (Volts - seconds/ Ampere - meter) In these equations there are four conditions that must hold at any surface of discontinuity(Young, 1952), which are: 1. The tangential component of E is continuous 2. The discontinuity in the normal component of D is a function of the surface charge density 3. The discontinuity in the tangential component of II is a function of surface current 4. The normal component of B is continuous Assuming fields having sinusoidal or time-harmonic dependence, phasor notation can be used for convenience. Thus, all fields will be assumed to be a complex vector 31 with an implied eja" time dependence. Therefore, the time derivative in the Maxwell’s equation can be replaced by the derivative of ejw‘ as shown in Equation (3-2) 3 jar . jar — e = 160(e ) (3-2) The Maxwell's equations can be written in phasor form by substituting Equation (3—2) for the time derivative term in Equation (3-1), and suppressing the e)“ term to give Equation (3-3). So the time derivative in the Maxwell's equations is replaced by jtu as shown in Equation (3-3) . VxE= -ja)B VOD=p D=£E (3'3) VxH=ij+J J=O'E VOB=O B=uH 3.2.2 Maxwell’s Equations in a Source Free Homogeneous Region Assuming wave prOpagation along the z-axis in a source free, linear isotropic, homogeneous region, the Maxwell's equations can be rewritten from Equation (3-3) as shown in Equation (3-4). Note that these assumptions imply that the electric charge density and the current charge density are both equal to zero. In writing these equations the flux density quantities, B and D are replaced by the field strength quantities as defined in Equation (3-3). 32 VxE =—jaer VxH=jmeE VOB=0 VOE=0 (3-4) 3.3 Electromagnetic Fields in a Cylindrical Region 3.3.1 Empty Cavity Solutions This form of Maxwell’s Equations (34) is the most useful form, from which modal solutions can be derived. But, first a coordinate system must be defined since modes are solutions to Maxwell’s equation in a defined structure. A cylindrical coordinate system Figure (3- 1) is defined since the solutions are sought for a cylindrical cavity. The two curl Equation (3-4) can be further decomposed into the vector, transverse, field z-components in cylindrical coordinates as shown in Equation (3—5). In these equations the z-components on the right side of the equations are the solutions that will be sought 33 :C D Figure 3- 1 Cylindrical coordinate system 34 2 2 E .. 1%.,- 1 a E: H =[_,-_1_§_Iz..11§§. P e do pate apt): P pate apdz u p 341 E- 13112 31 132Ez H =_ 3 1 laznhlasz f 63p ’uwepapé‘z f Mariette/118p (3'5) ' .1 82E 2 .1 32!! 2 Ez=—JLE[—a_z_21+fi 51! [if-111747” Hz] The two curl equations are simplified into the form of the Helmholtz equation by solving for either the E or H component in the two curl equations, then taking the curl of the electric field (or the magnetic field) curl equation as shown in (3-6), and substituting the curl identity in (3-7) to give (3-8). VxVxE= -j(oquH = (-ja#t)(-ja£E) 0-6) = cozueE V x V x A: V(VoA)—V2A (3-7) VZE + (0211313 = 0 (3-8) VZH + (02de = 0 Equation (3-8) is the wave equation or the Helmholtz equation which, with the apprOpriate boundary conditions yields an eigenvalue problem whose eigenfunctions are 35 Bessel’s functions and harmonic functions. These eigenfunctions describe the electric and magnetic field components of a mode while the eigenvalues, index the modes and describe the propagation characteristics of the mode. In solving these equations for a cylindrical cavity the following assumptions are used. Additional assumptions are used depending on the electromagnetic mode which the equation is used to describe. Assumptions are(Pozar, 1991): l. Tangential component of the electric field is equal to zero at the cavity walls, i.e. E¢(r=a,¢,z)=0 2. Fields must be finite everywhere, i.e. Bessel’sl’s function of the second kind cannot be a-solutiOn since, when the argument becomes zero the Bessel’s function of the second kind becomes infinite. 3. Fields must repeat every 2 pi radians in (it 4. The electric field components at the tOp and base of the cavity are equal to zero, i.e. Ep=0, (z=0, z=h) and E¢=0, (z=0, z=h) 3.3.1.1 Modes There are two different types of modes that can propagate in a cylindrical cavity, TE(transverse electromagnetic) or TM(transverse magnetic). TM modes are solutions to the Maxwell's equations with the boundary condition that there are no longitudinal magnetic field components while the TE modes are the solutions with the boundary condition that there are no longitudinal electric field components. In other words, in TE modes the electric field is aligned perpendicular to the direction of wave propagation and in TM modes the electric field is aligned in the direction of wave propagation. 36 33.1.1.1 Transverse Electric Modes - TE Modes For TE modes the electric field components are transverse and the magnetic field components are parallel to the direction of wave propagation, which is the z-axis and thus Ez=0. The field components from Equation (3-5) is reduced to Equations (3—9)(with the substitution of Ez=0), where a solution to Hz is sought in the Helmholtz Equation (3-10). 18H 1 32H E _-_ .___2. H = _'____;. P {83¢} P (Juaredez+) 13H 1 132H E = ——1 H = ' 1 (3-9) “6310] ¢{1#a£pa¢az] l 32H 2 E = H =-' —5—+ H z 2 133—3334 332 3 z] V2113 + 13193 = 0 (3-10 Using a vector potential function and separation of variables, a solution can be Obtained with the apprOpriate boundary conditions to give Equation (3-11) (Harrington, 1961). X 910,412) = anme( 2'" r)[sin(m¢)]sin(%z) m=0,1,2,3.... (3-1 1 nn=LL3 ..... where the TE mode field components are given in Equation (3—12). 37 "lit: .. __ Ep - p 3333 E1 J 333x” mp)sin(m¢)sin(ph flz) 5333:9321 =EZJ’(X Zmp)cos(m¢)sin(—z) Ez=0 . =0 1 131,1 = :11] ’X2'"( )cos(m¢)cos(—z) p #939 1'" p . l 132 X . 1r ¢=—j (0118 p 333; = Hsz(——am—"-p) srn(m¢)cos(-ph—z) H =- j l azwfazty W+pzw= me( ’annp)cos(m¢)sin(p—z) (3-12) 2 01183982! 322 2 _3(D[JH3am 30111130 H_H3a(_pl)3 H30 —_m(r p7!) El” an p’ 52:]. an ’Hl’xmn h H2: an 2 h 3.3.1.1.2 Transverse Magnetic Modes-TM-modes For TM modes the electric field components are parallel and the magnetic field components are transverse to the direction of wave propagation which is the z-axis, and thus Hz=0. The field components from Equation (3-5) is reduced to equations (3-13), where a solution to E2 is sought . 325 (1135] Hp: —-—-=- 1W!uwe3p5‘z! #1989 2 1 185 185 E =' l H = ———Z- 4’ [111016123932] 4’ {II 310) (3'13 2 52" . 1 {31:2325 333] Hz: #916182 for 38 Using similar boundary conditions as in the TB modes solution, a final solution is derived for the TM mode field components in the form of a vector potential as shown in (3-14). X ‘1’(P,¢.z)= Amnp1m(—;3"E-P)COS(m¢)COS(-€Ez) m=0,l,2,3.... (3-14) p,n = 1,2,3 ..... Where the TM mode field components can be simplified as shown in (3-15). 2 X . 33 = “1.3115313; = 13,];. (7:45p) cos(m¢)srn(£h£z) 2 E, = - 1313-3313-3333- = E21..(-{:"-p)sin(m¢)sin(£h£z) 2 E2 = -j—1-[§—ZI-+ 3.21;!) = E3J333 (Eflp)cos(m¢)cos(££z) rope & a h -1122 = I... - 2.7: H33 - [1 p 84> H,J,,( a p)srn(m¢)cos( h z) "12$ = , Ea £1 H3 .. 33 3p H214 a p)cos(m¢)cos( h z) (345) H3“,o £32931) iii a at} beat“: :2 peas“ E‘ x", h 5’ X2m(p1h H‘ 1X2... p H2 1X,“ 3.3. 1.2 Modes Designation The important information embedded in these Equations (3-12) and (3-11) is the characteristic field patterns associated with these modes. These field patterns indicate the variations and amplitude of the field components as a function of the cavity axis. In the cylindrical cavity the field indices (m,n,p) correspond to the components (¢,r,z). Where r, is the radial direction, (it is the circumferential direction, and z is the vertical direction. 39 Note that the order of the indices ¢,r,z is not in the cyclic order of coordinates r,¢,z, because it is common in circular waveguides to designate the 1p variation by the first subscript(Ramo, 1953 ). The index in is the periodicity in the radial direction, n is the number of half wavelength in the circumferential direction, and p is the number of circular wavelengths along the vertical axis. For a example, Tanp or TM() 10 indicates a transverse magnetic mode with (0) wave variations along the radial direction, (1) wave variation in the circumferential direction, and (0) wave variations in the vertical direction. 3.3.1.3 Electric Field Pattern The field patterns for the different modes can be determined by plotting the magnitude of the electric or magnetic field components. It is only necessary to plot these patterns for one plane since similar patterns are repeated in all the repeating planes along the z-axis. Several plots of the electric field patterns were generated using Mathematica and are shown in Figure 3-2 and Figure 3-3. See Appendix D for Mathematica program for generating these plots. These plots are shown as density plots where the white regions represent high field strength regions and the dark areas represent low field intensity regions. In generating this plots it was noted that for the TE modes, the 0 component was the predominant component while for the TM modes the 2 component was the predominant component. This is consistent with what would be expected for the different modes. 3.3.1.4 Cut-Off Frequency From the eigenvalues of the solutions to the Helmholtz Equation (3-9), the wave propagation characteristics can be derived in a form of what is known as the cut-off frequency equations. The cutoff frequency defines the minimum frequency for a given 40 cavity diameter where modes can prOpagate. The cut-off frequency equations for the TE and TM modes are shown in Equations (3—16)and (3-17). They show a relationship of frequency as a function of cavity length, h, cavity radius, a, permittivity and permeability, and the tabulated zeroes of the Bessel function. For a fixed cavity radius and microwave frequency the Waves of frequencies below the cutoff frequency of a particular mode cannot propagate, and power and signal transmission at that mode is possible only for frequencies higher than the cutoff frequency. At frequencies below the cutoff frequency of a given mode the prOpagation constant is purely imaginary, corresponding to a rapid exponential decay of the field and the generation of evanescent modes (Pozar, 1991). I 2 2 0’)” = 1 (“m") {1%) "mp 231/118 a h (3- 16) x333" = tabulated zeroes of the derivative of the Bessel' 5 function 2 2 TM = 1 xmn (pit) (”mp 222% [ a )+ h (3'17) x", = tabulated zeroes of the Bessel's function The significance of the cut-off frequency equations is that they indicate the frequencies and cavity lengths where a single mode can resonate in a cavity of a known radius. Another useful form of these equations is a plot of’the frequency versus the cavity length for a fixed cavity radius, to generate what is known as the mode chart (see Figure 3-3 and Figure 3-4. The mode chart shows the locations of modes with respect to other modes as a fucntion of frequency and cavity radius. It is important to note that these equations can be used in two forms: by fixing the frequency at a contant value and 41 varying the cavity length to excite different modes or by ; fixing the cavity length at a constant value and varying the frequency to excite different modes. The previous method is the fixed frequency mthod which is used in this work and the latter is the variable frequency method. In the variable frequency method more modes can be theoretical excited than in the fixed frequency method. This is evident by the number of modal lines that a vertical line intersects for the variable frequency method and the number of modal lines that a horizontal line intersects for the fixed frequency method. By in5pecting the mode chart it is apparent that, as the order of modes increase(increase in the indices) the modes become close together in frequency. Thus, suggesting that it may be difficult to excite higher order single modes. The mode chart also indicates that the modes that can be excited at a fixed frequency. TE(11x) TE(21x) TE(3lx) TE(13x) TE(02x) J I TE(61x)IA I Figure 3- 2 Electric field pattern for TE-modes TM(le) TM(llx) TM(12x) TM(le) TM(32x) Figure 3- 3 Electric field pattern for TM-modes e is": as We 3 a tee 1217.91.13 1‘.’ \IEflJ 112 fi _ . ' \\ 15012 h it (21.1) Aouenborr Figure 3- 4 Mode chart for TE-modes - Cut-off frequency versus cavity length 30 35 25 20 caiviy length (001) 15 10 sex i. l 4 LII a $5 i % “ewes-3 2 E: E $ EEEEEE33 :8 In N 8e 8. 5 U) s > .‘E 98 .‘n 4. (2H) Aouenbar; Figure 3- 5 Mode chart for TM-modes- Cut-off frequency versus cavity length (A) sic mic 11101 theo mere Voulr. 5181611 1when 1961) 1 hr min) 46 In general mode has a unique resonant cavity length, resonant frequency and field pattern associated with it However, there are TE and TM modes where the field patterns are different but the resonant frequency or cavity length are identical. These are known as degenerate modes and can exist at the TM-mode index of (1 lz) and the TE-mode index of (012), where the z-index can be any integer greater than 1. Mathematically, this can be seen from the cut-off frequency Equations (3—16) and (3—17), which shows that for a given cavity radius and resonant frequency, this can happen when the root and derivative of the root of the Bessel function are equal. It is also apparent from (3-16) and (3-17) that, as the indices or order of modes increase for a given cavity radius and resonant frequency, the resonant cavity length increases. This is logical since more half wavelengths are available with the higher order modes and thus, a longer cavity length would be required. 3.3.1.5 Cavity Quality Factor The Q-factor of a resonant cavity is defined as the ratio of the energy stored in side the cavity volume to the energy lost to the cavity surface area per unit time Equation (3-18). The Q~factor of a cylindrical cavity is a function of the resonant mode and in the microwave frequency range is usually very high and ranges from 10,000 to 40,000 or more(Collin, 1966). At a given frequency, as the order of modes increases the theoretical Q increases. This is logical since for higher order modes the cavity volume ' increases, and there is a greater volume-to-surface ratio, and energy is stored in the voulme, whereas it is lost on the imperfectly conducting surface(Collin, 1966). In practice the Q-factor can be lowered by the introduction of a feed system(coupling probe), a large impedance, imperfections in the construction, and imperfect conductivity at the cavity surface(Ramo, 1953 ). Equation (3-18) (Hanington, 1961) is the general mathematical definition of the quality factor. In this equation R is the intrinsic wave resistance of the metal walls. 47 snazdt energy stored = —‘i—— 3-18 energy lost ) ”fa RflHIst ( ) s Q=2ufo[ The importance of the Q-factor in processing is that it provides a guideline to how well a sample would heat at a particular mode, which is indicated by a decrease in the cavity quality factor as it is loaded with a sample. The theoretical Q-factor equations are only valid for ideal cavities where the internal surface of the cavity is homogeneously smooth, and for sample loads that do not perturb the theoretical equations significantly. However, in practice the sample loads are typically very lossy and large which invalidates the use of these equations because they are derived from perturbation theories. Thus, an experimental method using the bandwidth of the half-power points of a power response curve is used to calculate the cavity Q as indicated in Equation (3-19) and depicted in Figure 3-5. The power response curve is a trace of the frequency (x-axis) versus absorbed power (y-axis) that is generated on an oscillosCOpe using the low power cavity diagnostics technique(see appendix B). energy stored ___ 2 _f_Q_ (3- 19) energy lost Af Q=2nf0 1 -P — — - Power 2 ,, l 1 )3 frequency Figure 3- 6 Q-curve calculation using half power point method 3.3.2 Loaded Cavity Solutions 3.3.2.1 Small- Low loss Samples The modal solutions discussed above are for a homogeneously loaded cavity or an empty cavity in this case. There are other solutions for a cavity loaded with a small sample or a sample with a low dielectric constant using perturbation techniquesaianington, 1961). In order for the perturbation technique to be applicable it is assumed that the modal fields in the perturbed cavity can be approximated by those in 3 the empty cavity. Hence, the actual perturbation due to a sample load must be less than one percent, which is measured by the change in the resonant frequency or the change in energy in the cavity. For a composite material the dielectric constant of the polymer is approximately four, which is a relatively large value that would perturb the fields beyond the perturbation limits. Additionally, the size of composites that are processed are typically large and would perturb the fields beyond the perturbation limits. Hence this method cannot be applicable for a composite material load. C! it it 1 C01 dist but Very [it e. 49 3.3.2.2 Homogeneous and IsotrOpic Samples For a cavity loaded with a homogeneous material of various dielectric properties and load configurations, a mode matching technique was used to solve for the field equations in the cavity(Mannering, 1992 ; Lin, 1989 ). The load configurations were 3 classified as the cavity short-type, cavity Open type, and cavity image type(Mannering, . 1992 ; Lin, 1989 ). Solutions were derived by requiring that the modal field components at the boundaries of the loaded and unloaded regions of the cavity be continuous, or matched, and thus the name mode matching. Again, these solutions cannot be applied to a composite load because of the anisotropic and inhomogeneous characteristic of composites due to the combination of fiber and matrix. It is very important to note that the coupling of the modal fields with an anisotrOpic material like a composite, is so radically different from that of isotropic and homogeneous materials that the solutions cannot be linearly extended. 3.3.2.3 Simplified Composite Material Wei(Wei, 1992) used a simplified five parameter model to solve for the fields in a composite material for a TM(transverse electromagnetic) propagating wave. In this solution it was assumed that the incident wave on each composite surface was a linearly polarized TEM(transverse electromagnetic) wave. This means that the electric field vector points in a direction where the magnetic field will be ninety degrees away from it. It was also assumed that the power dissipated in the composite from the edge of the composite was constant through the thickness of the composite, and thus the electric field distribution inside the composite was only due to the propagating waves from the top and bottom of the composite. These simplifying assumptions were made as a first approximation for solving a very complicated problem. In the actual coupling between the composite material and the electromagnetic fields, the power dissipated in the composite is not constant through C011 5101’! 10 [h the d 50 the thickness of the composite, and there is a difference in coupling between the incident waves from the different planes of the sample due to the fiber reinforcement. Although, these solutions are good approximations to the exact solutions they are only valid for special cases. For composite materials the solutions are far more complex and computationally intensive. Hence, in the microwave processing of composites empirical relationship have been developed to aid in the understanding of the process. In the sections that follow concepts of microwave porcessing in a single-mode resonant cavity are discussed. 3.4 Fundamentals of Microwave Heating 3.4.1 Microwave Power Absorption and Perrnittivity To this point, fundamentals of electomagnetics have been discussed in terms of modes and modal field solution. Now the interaction of these fields with materials and how they heat will be discussed. The material property of non-magnetic materials which determines its ability to absorb microwave energy is complex permittivity is shown in Equation (3-18) below. if = e’-js” 0' (3-20) sow 8” = 8”+ The complex permittivity is mathematically a complex quantity which has two components; a real part which is called dielectric constant, and is related to the energy stored in the material; and an imaginary part which is called the loss factor and is related to the energy dissipated as heat in the material. The loss factor is due to contributions by the dielectric loss factor or the motion of dipoles and the conductivity or the motion of 3.4 ‘ mile] 1'8 51 charges. It is a function of material structure, composition, angular frequency, temperature, and pressure. As a dielectric material is exposed to an electromagnetic field, the polar segments attempt to align up with the alternating electromagnetic field so that the normal random orientation of the dipoles become ordered. These ordered polar segments relax and oscillate with the fields. The energy used to hold the dipoles in place is dissipated as heat into the material while the relaxation motion of dipoles is out of phase with the oscillations of the field. Materials with low dielectric loss do not absorb microwave energy as readily as those with higher dielectric loss. The absorption of electromagnetic energy by materials is not only dependent upon the permittivity, it is also dependent upon the angular frequency of the microwave and the square of the magnitude of the electric field strength inside the material as described in Equation (3-21). 1 ,, ‘2 Pz—Z-(oeoe |E| (3-21) Where: P = Microwave Power absorbed 8 = Permitivity of free space 0 to = angular frequency II 8 = dielectric loss factor E = electric field 3.4.2 Penetration Depth The angular frequency is typically constant and determines the depth of penetration of the electromagnetic wave into the material. Typically, as the frequency decreases the depth of penetration increases as the free space wavelength increases. A the incl in a 1mm table and: L“ 5ng 531115 33 isotropic mUSl Ca} 52 linear dimension called skin-depth or penetration depth is described as the depth at which the electromagnetic energy would decay or attenuate to lle. The parameter can be used to select the appropriate frequency and sample size so that volumetric heating is achieved. To achieve volumetric heating the power generated in the material over the skin-depth must be substantial compared with the actual dimension of the sample(Binner, 1991 ). Equation (3-20 ) can be used to calculate the penetration depth of various polymers and composites. Using this equation the skin-depth has been calculated for several polymers and composites at two different frequencies, 2.45 GHz and 0.915 GHz and summarized in Table 1. These two frequencies were chosen because they represent the two frequencies that are widely used. They also represent current Operating systems in our lab which are used in a scale-up study of a 7 inch cavity to an 18 inch cavity to process 3 inch and 8 inch samples, respectively. Typically, in order to excite similar electromagnetic modes in a larger radius cavity, the frequency has to be decreased. This can be better illustrated from the cut-off frequency Equations (3-16) and (317). According the values in the table, at these frequencies substantial penetration depths can be achieved for the polymer and composite systems, and thus volumetric heating should be achieved. L flou’b‘ob" (3-22) In Equation (3-22), Z is the penetration depth in meters, and the other variables are the same as were previously defined. It is important to note that penetration depth is an isotrOpic concept. Hence, in order to calculate it for the anisotropic composite material it must calcualted for the principle axis of the composite. For the AS4/3501-6 composite, 53 the effective permittivity along the fiber direction and perpendicular to the fiber direction are given by (Lee and Springer 1984) to be 1- j 2500 and 14.5 - j 75.8, respectively. Assuming the fiber direction to be the principal axis, the skin depth is calculated as shown in the Table 3-1. This table shows that at 2.45 GHz the penetration depth for several polymers can range between 0.4 m and 88 m; and between 0.0098m and 0.0032m for a composite. This table also indicates that at a lower frequency, 915 MHz the penetration depths are higher as compared with those for the 2.45 GHz values. Material e/eo Z(m) Z(m) £2.45 f=915 GHz MHz VfiyTester 3.44 - j 0.18 0.7 1.08 Tilly cured gioxy 3.5 - ' 0.1 0.73 1.95 Uncured epoxy 4.25 3' 0.25 77.32 0.86 Fiberglass 4.4 - ' 0.13 0.63 1.68 olystyrene 2.55 -j 0.0MSS '73 196 PolyethLleneLpure) E25 - ' 0.0007 83 223—“— AS4/3501-6 l-j 2500* 0.01 0.03 Graphite / Epoxy 3 com osite(alon fiber) EFL—$435014 14.5 - j TT—s. * '3': 8.5 Graphite / Epoxy composite(perpendicular to fiber) *effective perrnitivitty Table 3-1 Penetration Depth for Samples at various frequencies 3.4.3 Applicators 3.4.3.1 Multimode and Waveguide There are several types of microwave applicators with benefits that range from low cost to efficient energy coupling. Some of the different applicator types that have been used in the processing of materials are waveguides, multimode cavities, and single- mode cavities(Asmussen, 1987 ). The waveguide applicator is the easiest to fabricate and is mad up of a hollow pipe. The multimode applicators comprise of over 50% of all 54 industrial microwave applicators and is based upon the principle of the commercial microwave oven(Asmussen, 1987 ). In a given frequency range the multimode applicator is designed to support a number of resonant modes. Thus, it reduces coupling sensitivities since several modes share heating and compensate for load prOperty changes(Asmussen, 1987 ). To improve heating uniformity mode stirrers or the multiple generators are used to distribute the power and provide for better exception of the modes(I-luack, 1969 ). The wide use of this applicator is due to the low cost, ease of construction, and the adaptability to a wide variety of material loads of different sizes and effective loss factors(Asmussen, 1987 ). 3.4.3.2 Single-Mode Cavity Unlike the multi-mode applicator, the single-mode applicator is very sensitive to load changes, but most efficient in the electromagnetic energy coupling. A single mode resonant cavity is made up of a hollow pipe and two plates for top and bottom. The top plate is adjustable similar to the adjustment of a piston while the bottom plate is fixed. In the applicator studied the microwave energy is coupled to the cavity by an antenna which is inserted horizontally through the side of the cavity wall . At a given frequency the applicator is designed to support a single mode. As material load properties change during heating the cavity has to be adjusted to compensate for these changes. The adjustment to compensate for these changes is called mode tuning. For mechanical tuning, the cavity volume is adjusted by the movement of the cavity top and the depth of the coupling probe. For variable frequency tuning the cavity length and probe depth are fixed as the frequency is adjusted to compensate for material load changes. These tuning relationships can also be observed from inspection of the cut-off frequency Equations (3-16) and (3—17). 55 3.4.4 Mode Tuning The method of electromagnetic energy coupling and matching used in the single mode resonant cavity is similar to that used in ion and plasma sources(Asmussen, 1974; Asmussen, 1984 ; Asmussen, 1985; Asmussen, 1986 ) and electro-thermal engines(Whitehair, 1984; Whitehair, 1987 ) The fundamental of this method can be described from the principles of RLC circuits where the input impedance, Zin of a microwave cavity is given in Equation (3-23)(Asmussen, 1987 ). P, + j2ro(Wm — we) 3 = = RIF! + inn (3'23) 1 2 _llo! 2 Zin In this equation, P1 is the total power coupled into the cavity including losses, Wm and We are time-averaged magnetic and electric fields stored in the cavity fields, reSpectively, llol is the total input current on the probe, and Rin and inn are the cavity input resistance and reactance or the complex load impedance as seen by the feed transmission line. 'At resonance the impedance, Zin is minimum and is equal to the resistance, Xin as shown in (3—24). Above or below the resonant frequency the impedance is maximum and is expressed in Equation (3-25). 2. = R- (324) Zin = Rn. +J'Xn. (3-25) Thus, to achieve resonance two independent adjustments are required to match the cavity load to the transmission line. One adjustment must cancel the load reactance, 56 Xin while the other must adjust the resistance, Rin to equal to that of the characteristic impedance of the feed transmission(Assmusen, 1987). The cavity length and probe depth are two variables that are used to achieve this goal. At resonance maximum energy is focused into the cavity. As the cavity is loaded, the energy that is focused into the cavity becomes a function of the resistance of the load. As the resistance of the load increases, more energy is lost to the load and the energy stored in the cavity decreases. This is characterized by the flattening of the power response curve and the decrease in the cavity quality(Q) factor. 3.5 Summary In this chapter the fundamental concepts of electromagnetics and microwave processing were discussed. Fundamentals of electromagnetics were presented from the point of view of Maxwell's equations and how they relate to microwave processing. This was done by the simplification of the Maxwell's equations to generate the wave equation which, was then solved as an eigenvalue problem with apprOpriate assumptions and boundary conditions for a cylindrical cavity.. Fundamental concepts of microwave processing was discussed from the point of view of theoretical processing issues such as loss factor, microwave power absorption, and skin depth. Other processing issues such as applicator challenges, and applicator characterization as in impedance matching or mode tuning and quality factor calculations were also discussed. In summary, this chapter should address all relevant theoretical electromagnetic and microwave processing issues that will be used in this work. in CO CHAPTER 4 EMPTY AND LOADED CAVITY CHARACTERIZATION 4.1 Introduction A key in the microwave processing of composites is the understanding of the electromagnetic fields behavior inside the cavity. The magnitude of the electric field strength determines the heating rate and the field pattern determines the heating uniformity. These field characteristics are a function of the sample dielectric properties, size, geometry, fiber orientation, placement, resonant frequency, and resonant cavity length. Thus, in order to control the heating inside the cavity the magnitude and field pattern of the electric field inside the cavity must be known. For each electric field pattern there is a correSponding electromagnetic mode. In the single-mode resonant cavity the measurable parameter for identifying an electromagnetic mode at a fixed microwave frequency is the cavity length. As a sample is loaded into the cavity, the known empty cavity lengths corresponding to the electromagnetic modes become altered(see Chapter 3) and mode identification using cavity length becomes complex. Because the electromagnetic modes are discrete, a fundamental mathematical approach for identifying loaded cavity modes was found to be computationally intensive. Using graphite! epoxy, glass / polyester and un-reinforced nylon samples experiments were done to understand the electric field behavior in the loaded cavity. This was done by measuring the radial electric field along the axial and circumferential axis, and the resonant frequency shift of the loaded cavity at a TM(012)-mode. 57 Ca dis C0“ 58 Graphite / epoxy and vinyl / ester glass composites were then heated to identify heating characteristics of the loaded cavity modes. But first empty cavity solutions were used to characterize the empty cavity, in order to define the actual empty resonant cavity lengths and quality factors in the presence of circuit irnpedances and cavity imperfections. 4.2 Empty Cavity Characterization 4.2.1 Experiments With the theoretical cavity lengths available from the cut-off frequency equations(Equations 3-16 and 3-17), the empty cavity modes were measured using the swept frequency method. This was done by adjusting the cavity length around the theoretical values in addition to adjusting the coupling probe depth to achieve resonance in the empty cavity, which was measured by a power absorption trace on an oscillosc0pe. Adjustment of the cavity length aligned the power absorption curve along the excitation or center frequency and the adjustment of the coupling probe increased the amplitude of the curve along the y-axis. The power absorption curve was then used to calculate the cavity quality factor or Q-factor as explained in Chapter 3. This procedure was repeated until all of the observable theoretical modes were measured. 4.2.2 Results and Discussions - Empty cavity characterizations Table 4-1 lists the calculated cavity lengths, the corresponding measured empty cavity length, probe depth and cavity quality factor. Of the 1? calculated modes, 15 were observed which include 10 TE-modes, 5 TM-modes and 2 degenerate modes. As discussed in Chapter 3, TE-modes are electromagnetic modes where the electric field component is transverse to the direction of propagation, and TM-modes are those where 59 the electric field component is parallel to the direction of propagation. In these designations the vertical axis is taken as the direction of prOpagation. The degenerate modes were observed at the measured cavity lengths of 11.08 cm and 22.46 cm. Also, as mentioned in Chapter 3, degenerate modes are TE and TM modes with different field patterns but with the same resonant frequency or resonant cavity length. In practice cavity imperfections, circuit impedances and the presence of the coupling probe inside the cavity could all contribute to increased energy losses in the cavity which results in a decrease in the cavity volume required to support the corresponding calculated mode(Mannering, 1992). Thus, measured cavity lengths are typically lower than the calculated ones. In this work the measured cavity lengths were approximated to the calculated ones by assuming that they correspond to the closest values that are greater than the measured values. For example, a measured cavity length of 11.08 cm was approximated to the calculated value of 11.283 cm and not 13.383 cm as shown in table 4-1. Higher order modes corresponding to TE(114) at 26.767 cm and TE(214) at 32.971 cm were not observed. Typically, as the order of modes increase (the increase in the mode indices) the wavelength decreases, the resonant cavity length increases and the modes become very close together in frequency(sec mode chart in Chapter 3). As the modes become closer together in frequency they become difficult to excite . independently. Hence, the un-observed modes could have existed as a hybrid of other higher order modes that were excited in this cavity length range. It is important to note that the un-obcrscrvable modes were both TE-modes that were one 111 variation apart. Modes Theoretical Measured Probe Cavity Cavity Cavity Depth Q-Factor Length Length (cm) (cm) (mm) (K) m 13.383 1319 5.64 18 WW2), 14.41 14.39 1277—. ‘7'— “TEE—N 11) W . 1 _'6—15.1 175—“. ”217—" W) 16.485 16.44 4.09 8 WW7015 7.66 8 m(or3) 21.615 2T.7"'7—TBS 3 Wham 22.48—70.04 9 mm— 2.48 ”20.04 9 21) WW'TIG 4 WW WERE—"27.01 .21 4 12) WW1 3.4T 9 WW1 TE(tts) 33.459 33.29 4.87 '23—— Table 4- 1 Measured and calculated resonant cavity lengths The corresponding cavity Q-factors are also shown in Table 4-1 which range from 3000 to 23,000. The Q—factors did nor follow a specific profile by increasing as the order 61 of modes increased as was expected from the calculations. Again this could be due to the characteristic impedance of the cavity and the circuit due to the imperfections in the cavity as mentioned before. As discussed in Chapter 3 the cavity quality factor is a ratio of the energy stored to the energy lost in the cavity, and therefore as the energy loss to the cavity increases the Qfactor should decrease and the reverse is true. Typical Q factor values can range from 10,000 to 40,000 or more for ideal cavities (Collin,l966), where ideal means perfectly and homogeneously smooth and perfectly conducting cavity walls. The measured Q values can fall in any range and can serve as a fingerprint of the coupling efficiency of the cavity and circuit. In processing it can be used as measure of the coupling efficiency between sample loads and microwave energy inside the cavity. In other words it can be used to determine how a sample would heat at a particular mode. A decrease in empty cavity Q or low Q value is an indication that microwave energy is being lost to the sample load, or the sample load is efficiently absorbing microwave energy inside the cavity. In this work loaded cavity Q will be used to select the most efficient heating modes. The lower the Q the more effective the coupling and heating would be. 4.3 Loaded Cavity Characterization 4.3.1 Experiments Experiments were performed using nylon (dielectric constant=2.06), polyester] glass composite (dielectric constant=4.50) and graphite / epoxy composite(dielectric constant = 4.23) to identify the loaded cavity characteristics of a TM(012)-mode. All samples were cut into one inch squares, which was the size calculated from perturbation theory to have minimal field perturbations for a material with a dielectric constant of 4.0. Two different sample heights were used, one was 1/8th of the free-space wavelength 62 away from the coupling probe which will be referred to as the lowered sample position(free space wavelength for 2.45 GHz is 12.24 cm) and the other was a quarter of the free space wavelength away from the coupling probe which will be referred to as the elevated sample position. The choice of these sample heights was in accordance with typical processing protocol such that maximum cavity volume is available for mode excitation. A solid Teflon block was used to elevate the samples to the appropriate heights. For the lowered sample position a square 7.81 cm in length by a 1.88 cm thick Teflon block was used to elevate the samples from the base of the cavity. For the elevated position, the square block was placed on t0p of another solid Teflon disk that was 12.25 cm in diameter by 1.88 cm thick. The one inch square samples were placed in nine marked positions, one at a time such that it spanned a nine inch square area which represents the area of a typical three inch squared sample. For each sample placement the cavity was Opened and the sample was physically moved and carefully placed in the specified location. 0.882" 0.882? 2 3 4 9 1 5 8 7 6 Figure 4- 1 Sample Placement mC cm, by z prof 4.3.1 deter. diagn was 11 . “my referrer £73th {05 the 63 The samples were placed one at a time where the order of sample placement followed the numbered order in Figure 4-1, with the center position being the first placement. The coupling probe position was always close to position 7. When using composite materials, the fiber orientation was referenced to the coupling probe placement. For example, a perpendicular fiber orientation would mean that the fiber alignment in the unidirectional composite material was perpendicular to the coupling probe. Positions 2 and 4 are referred to as t0p left and t0p right, respectively and positions 6 and 8 as bottom right and bottom left, respectively. In all of the measurements, at the resonant empty cavity length for a TM(012) mode a sample was loaded into the cavity which caused the resonant conditions of the empty cavity to change. The loaded cavity was then tuned to the new resonant conditions by adjusting the cavity length and probe depth. At the new resonant cavity length and probe depth the measurements were done to identify the characteristics of the mode. 4.3.1.1 Frequency Shift Measurement In the frequency shift measurement a reference resonant frequency was first determined for a TM(012) mode, by tuning the cavity to resonance using the low power diagnostic system to generate a power absorption curve on an oscilloscope. A sample was then loaded into the cavity and the coupling probe was adjusted to fine tune the cavity by increasing the amplitude of the power absorption peak By assuming that the reference resonant frequency was 2.45 GHz, the frequency shift was measured by graphically determining the difference between the reference resonant frequency, and that for the loaded cavity on an oscilloscope trace as described in appendix B(see Figure 4- 2). ' 64 Because a Teflon block was used to elevate the samples, it was used to set the reference resonant frequency and the frequency shift was determined from the difference between the resonant frequency due to the Teflon load and that due to the addition of the sample load. l /.\ I \ \ Power / ' \ 1 * )Js‘r, H: Inch-c: M-tnquency dam Figure 4- 2 Frequency Shift Measurement 4.3.1.2 Electric Field Measurement In the electric field measurement, a micro coax probe specifically designed for E-field measurement was connected to a power meter and inserted into the cavity through holes that were drilled through the cavity walls. The extension of the probe into the cavity was limited to 1.0 mm so that it caused minimal perturbation to the resonant mode ' field patterns and the resonant frequency during measurement. The power detected by the probe and measured by the power meter is proportional to the radial electric field inside the cavity or lEI’. The microwave cavity was designed with radial and axial E—field diagnostic holes specifically for accommodating the micro coax probe for the electric field measurement. The axial E-field diagnostic holes were located at the direct opposite side of the coupling probe and extended to the total height of the cavity(see Figure 4- 3). There were four rows of circumferential diagnostic holes which were placed at approximately 1.2 inches 65 from the base of the cavity, at approximately the same height as the coupling probe(see Figure 4- 3). These rows extend approximately 140 degrees or 40% of the cavity circumference due to the presence of the cavity support bracket. Ideally, these circumferential diagnostic holes should extend the full circumference of the cavity for accurate diagnostics. Support Bracket flm Coupling Probe I w I l WA WI]. Figure 4- 3 Electric field diagnostic holes: a) front view, b) side view, c) top view 66 25 ::: 26 ::: 27 ::: ‘ 28 :2: V Figure 4-3b Side View of cavity showing doagnostic E-field diagnostic holes 21 22 23 z 24 250000 0000 260000 0000 270000 0000 280000 0000 30——-> 031 4) l/ 32 \ Figure 4-3c Front View of cavity showing doagnostic E-field diagnostic holes 67 To do the electric field measurement, the resonant empty cavity at a TM(012) mode was loaded with a sample and then tuned again to compensate for the sample load. Six Watts of microwave power at 2.45 61-12 was then coupled into the cavity while the e- field measurements were done. Electric field measurements were done from the top of the microwave cavity and along the axial and circumferential axis. 4.3.2 Results and Discussions - Loaded cavity characterizations A key in mode identification is the differentiation between TE and TM modes, and the identification of the mode indices. The E-field measurement from the t0p was used to differentiate between TE and TM modes. Since the electric field component is parallel along the z-axis for the TM-mode and perpendicular with the z-axis for the TE- mode, the measured E—field from the t0p of the cavity should be a value greater than zero for the TM-mode and equal to zero for a TE-mode. In this case a value of 12.0 was measured which confirms the presence of a TM-mode as was expected. The E-field measurement along the axial axis was measured to identify the numerical value of the index 2, which is the number of half wave variations along the axial direction(see Figure 4- 3). The radial E-field along the circumferential axis was also measured to identify the numerical value of d), which is the number of half wave variations along the circumferential direction(see Figure 4- 3). For the TM(012) mode where (6:0, rr-l, z=2) there should be two wave variations along the z-axis, and no wave variations along the circumference of the cavity since the TM(012) mode is 4) symmetric. Figures 4-4 through 4-9 show the axial elecuic field measurements, Figures 4-10 through 4—15 show the circumferential radial electric field measurements, and Figures 4- 16 through 4-19 show the frequency shift measurements. For the purpose of clarity, these results have also been summarized in tabular form in Tables 4-1 through 4-3. The axial field measurements in Figures 4-4 through 4-9 show multiple plots for sample 68 placements in the different location in the fixed z-plane. The y-axis shows the electric field diagnostic hole location along the vertical axis of the cavity, and the x-axis is the electric field measurement. The base of the cavity is represented by hole #0 and distance between holes #0 to #21 is equal to a cavity length of approximately 13.0 cm. The coupling probe placement height is indicated by a horizontal dashed line along the y-axis between hole #27 and #28. The elevated and lowered sample placements were also indicated by dashed lines between hole #27 and #28 and between #29 and #30, respectively. Table 4-2 summarizes the axial electric field measurements in which column 3 and 4 indicate the number of high amplitude peaks and their location along the z-axis, respectively. The number of high amplitude peaks is equivalent to z, the number half wavelength variations along the z-axis. From table 4-2, the empty cavity axial E-field measurement shows two high amplitude peaks with the maximum at holes #29 and #24. This is consistent with what was expected since the cavity length was that corresponding to a TM(012). Similar axial electric field measurement was also observed for the Teflon block loaded at the elevated position as shown in Figure 4-5a For the Teflon block loaded at the lowered position only one fully developed high amplitude elecuic field was observed at the bottom region of the cavity, and a partly developed one in the upper region of the cavity at hole #22 as shown in Figure 4-5b. This lowered Teflon profile is similar to the elevated Teflon profile except for that, the location of the high amplitude peaks are at higher locations in the cavity. For the elevated sample, the peak in the lower region of the cavity is located within the Teflon block while for the lowered sample it is located outside the block region. 69 Radial Electric Field Pattern along Axial axis for Empty Cavity Coupling Probe 10 12 14 IEI‘Z 16 18 20 Figure 44 Axial Field Electric Field Pattern for Empty cavity (a) Elevated (b) Lowered 7O z-axia hole position Axial Radial E-fleld Pattern for Teflon Cavity Load at the Elevated Position Sample and Coupling Probe 10 12 14 16 18 20 IEIAZ (a) Axial Radial Electric field Pattern for Teflon Cavity Load at the Lowered Position 21 22 4L 23 4. r: .3 24 3 a 25 .t .2 26 2 o 27 j Coupling Probe g 28 +— —————————— N 29 Sample 30 .T ------------------------------------- 31 ‘ 0 18 20 lEI‘Z (b) Figure 4-5 Axial Field Electric Field Pattern for Teflon (a) Elevated (b) Lowered 71 Axial Radial E- field Pattern for Graphite Epoxy Cavity Load at the Elevated Position-perpendicular fiber direction Wet MW earns Mop right corner -5—Bonom rim cane -‘-Bottom left comer Sample and Coupling Probe 10 12 l4 16 18 20 IEI“1 (a) Axial Radial E- i'ield Pattern for Graphite E y Cavity Load at the Lowered Position-perpendicular tber direction 21 , .- g 22 + \_ \X _ +Carter Mop id! -X-'Top rid! l +Bottom rigll Wm left a-axla hole position N a Figure 4-6 Axial Field Electric Field Pattern for Graphite / Epoxy Perpendicular Fiber Direction (a) Elevated (b) Lowered 72 Axial Radial E-fleld Pattern for Graphite Epoxy Cavity Load at the Elevated Position-parallel fiber direction 21 I 22J ’ ’\ _ -i-Center -O—‘1'opieftca'ner b \\ . -X-Top righcorner “ottomrightcomer 23 1L - [o’- Honour lefterxmer 5" 24 4» R a 25 r. '5 26* ' Sample and f 27 T - . Coupling Probe Te -- \ -- ---------------------- — ---------- i . - r'e 28 l _\ . 31 4- __ ‘ . v # e O 2 4 6 8 10 12 14 16 18 20 IE|*2 (a) Axial Radial E-l‘ield Pattern for Graphite Epoxy Cavity Load at the Lowered Position-parallel fiber direction 21 22 -°—Center “up left corner -fl-‘I‘op rigix corner Worn riglx owner m left corner 23 24 25 z-axls hole position N Oi 27 Coupling Probe 28 _______________________________ 29 Sample 30 ———————————————— 31 ‘ 0 2 4 6 8 10 12 i4 16 18 20 lEl‘Z (b) Frgure 4-7 Axial Freld Electric Field Pattern for Graphite / Epoxy Parallel Fiber Direction (a) Elevated (b) Lowered 73 Axial Radial E-field Pattern for Nylon Cavity Load at the Elevated Position 21 +Center -O—'l'op left corner —X-Top riglx comer 22+ ‘“ -O—Bottomrighteorner Midtown" 7.; 4- Sample and —--—-- ——--—- z-axls hole position '6! 29 .. 30 .. 31 4 1 O 5 10 15 20 25 30 35 4O IEI“2 (4) Axial Radial E-fieid Pattern for Nylon Load at the Lowered Position 21 -D—Center -O-Top left corner -X-‘l‘op right corner +Bottom left corner Coupling Probe z-axis hole position 8 $8 '83 ti 8 13‘. i: 13 13 W a-e (b) Figure 48 Axial Field Electric Field Pattern for Nylon (a) Elevated (b) Lowered 74 Axial Radial E-field Pattern for Polyester Glass Cavity Load 21 at the Elevated Position Hater -'°—Top left cane -X-‘I'op riglx comer 22" -D—Bottomrightcome +Bottomieitemner 23 4 3 24 4» ii 2.25 .. 8 26 .3 Sample and a 27 ‘ $23 “ Coupling Probe Height fl ———————————————————————————————— 29 4 30 1 31 1 r r r 20 25 30 35 40 IEI‘2 (3) Axial Radial E-l'ield Pattern forPolyester Glass Cavity Load at the Lowered Position 2; Met Mop left easier -X-Top riglx corner a 23 i “—Bottom rigix corner +Bottorn left corner e :5 24 a 25 .2 26 a e “ 27 ..... ' \_ ' ______________ C: "Pinata“. - :5! 28 \\ ,, _ .......... lam; 31 _ % 4 4 6 8 10 l2 14 16 18 20 IEI“2 (b) Figure 4-9 Axial Field Electric Field Pattern for Polyester Glass (a) Elevated (b) Lowered 7S Radial Electric Field Pattern along Circumferential axis for Empty Cavity 20 18» 16- 14~ 1241. mi +025 +U25 +I27 +028 6... 441- 2.1 11 ; *- pmbe 0.9m . 2.01 mm Probe Depth - 19.28mm Circumferential hole locations(3,4,5,6,7) Figure 4-10 Radial Electric Field Pattern for Empty Cavity 9- axis Radial EoField Pattern for Teflon Cavity Load at Elevated and Lowered positions 20 . , 18 - 4425 45—32 1 16 ~- 1 14 ., -a—¢27 -I—l2 i 12 4 g go 10 -» i E 3 r : - 6 1 g 4 -+ l 2 fi 1 0 4 fiflA a e 3 : Elevated Sample Positions Lowered F1 gure 4-11 Radial Electric Field Pattern for Teflon load at Elevated and Lowered Positions 76 IEI‘Z 20 0- axis Radial E-Field Pattern for Graphite Epoxy Cavity Load at elevated position- perpendicular flber direction 18 r 16 .. r4 4. 12 4 +025 +02 +027 +02 . . . . . . . .' Gil-.4 . .' Center Top left Top right Bottom right Sample Positions Bottom left (a) IEIM 20 9— axis Radial E-Field Pattern for Graphite Epoxy Cavity Load at lowered position- perpendicular flber direction 18 -. 16.1. 144 124 10 4» on- d ”.9 +025 +02 *027 *2 Center Top left Top right Bottom right Bottom left Sample Positions 0)) FigUr'e 4-12 Radial Electric Field Pattern for Graphite/ Epoxy Perpendicular Fiber Direction (a) Elevated (b) Lowered 77 o— axis Radial E-Field Pattern for Graphite Epoxy Cavity Load at elevated position- parallel flber direction 20 1 18 .. —o—s25 —u—s26 16 4. l4 0 -g—s27 *2] 12 4. N 10 1- < E 8 6 T 4 a» 2 J. 0‘ ‘ ‘ , . ./.‘/‘:;‘ _- . -. : fl. 4 .‘3- Center Top left Top right Bottom right Bottom left Sample Positions (a) b- axis Radial E-Field Pattern for Graphite Epoxy Cavity Load at lowered position- parallel fiber direction 13 .. -o—szs -o—n 140 +‘27—H2 lEl“2 on Center TOP Right TOP Lt“ Bottom Right Bottom Left Sample Positions (b) Figure 4-13 Radial Electric Field Pattern for Graphite / Epoxy Parallel Fiber Direction (a) Elevated (b) Lowered 78 o-axis Radial E-Field Pattern for Nylon Cavity Load at elevated position 20 18 ” +125 +02 16 -- 14 4? +‘27 +‘2 12 .t N 10 0 i. E 3 ‘r 5 .. 4 4» ‘ . o g I»..L' ;A“... o._":.-fl—.; --'.‘.-.£' "".L.2; enter 0' °P g Bottom Left Bottom Right Sample Positions (21) ¢- axis Radial E-Fleld Pattern for Nylon Cavity Load at lowered position or S. F! Center Top Left Top Rigb Bottom Right Bottom Left Sample Positions (b) Figure 4— 14 Radial Electric Field Pattern for Nylon (a) Elevated (b) Lowered 79 o-axis Radial E-Field Pattern for Polyester Glass Cavity Load at elevated position 20 18 «r 4425 4—‘2 161* 14 4. *27 -I—02 12 .1. " 10 .. S. g 81. f. fl 6 4. fix, 4 4» 0 ~ . . Center Top Right Top Left Bottom Left Bottom fight Sample Positions (a) b—axis Radial E-Fleld Pattern for Polyester Glass Cavity Load at lowered position 20 - 18 —o—s25 -¢I—s2 16 14 +027 +112 12 g! 10 E 8 6 , It 4 l3 2 ' ' ' ‘ 0 - A . _ 7% Center T°P Right Top Left Bottom Right Bottom Left Sample Positions (b) Figure 4-15 Radial Electric Field Pattern for Polyester/ Glass (a) Elevated (b) Lowered 80 u ShiftofCavi Loadedwmt ' “remap“...mm 1...... . Em form(012)Mode 0.014 0.01167 0.009333 0.007 0.004667 0.002 333 0 0.002333 1 0.004667 Marcy Slill Dela 0.007 0.009333 0.01167 0.014 (a) FrequencyShiftofCavityLoadedwithGrafliiteE y at lowered Position oPerpendiarlx Fiber Ducting: for TM(012) Mode :::; 0.002 FrequencyShiftDeita 0.004 0.006 0.01 (b) Figure 4-16 Frequency Shift Measurement for Graphite I Epoxy Perpendicular Fiber Direction (a) Elevated (b) Lowered 81 FreqmcySuftofCavitylpadedW'thraphiteEpoxy atEevuedPosition-PualieiFiberD'nerxion forMOlZ)Mode 0.01 6 0.01 333 0.01067 0.000 0.005333 0.002667 0 =I=2éssz » 9/ 1 Registry Shin DeMGilz) lflQ" 0.01067 0.01333 0.016 (a) Frequency Shift chavrTyloaied with GraphrWEpoxy m Lowered Position -Pm'allel Fibs Direaion for M012) Moth 3 0.009 0.0075 0.006 0.0045 0.003 0.0015 0.0015 Frequency sum Deita(Gllz) 0.003 0.0045 0.006 0.0075 0.009 (b) Figure 4—17 Frequency Shift Measurement for Graphite / Epoxy Parallel Fiber Direction (a) Elevated (b) Lowered 82 Shiftof ' Loaded w‘ Ny 13mm for‘l'M(012)Morh 0.044 0.0352 0.0164 0.0176 WMWOM 0.0176 0.0264 0.0352 0.044 (b) Figure 4-18 Frequency Shift Measurement for Nylon (a) Elevated (b) Lowered 83 Siift ofCavny' 16.4.41 with or... a ated Positron for worm 0.0332 3 “W ./ \. 0.0237 1 0.01897 2% . filfl 0.02371 . 0.02846 \ / 7 0.0332 Frequency Shift Dela (Gila) o o (a) WSlu'ftofCavityLoarledwitthiyaterGlau theLowaedPoaiionfaMOlZ) Mode 0.005 0.004286 0.003571 0.002357 0.002143 0. 001429 0.0007143 ; 0 0 .0007 143 3 0.001429 0.002143 0.002057 0.003571 0.004286 0.005 Frequarcy Shift Demoliz) (b) Figure 4-19 Frequency Shift Measurement for Polyester] Glass (a) Elevated (b) Lowered 84 ample Sample Number of high electric field amplitude PeEE Height peaks for sample placement in the : Locations hole # Center Top op Bottom Bottom Left Right Right Left 'Empty Cavity -- - -- - -- -- 29 24 Mon Elevated - -- - -- -- 30 23 raphite/ Epoxy Elevated 2 2 2 2 2 30' 23 Perp. Fiber Graatfiriiel/= Epoxy Elevated 1 2 2 2 2 30 23 Par 0 i r olyester / Glass Elevated 2 2 2 2 2 36 23 N lon Elevated 2 2 2 2 2 30 23 eflon Lowered - -- -- -- -- 28 22 Graphite 7 Epoxy Lowered 2 l 2 1 1 29 22 Perp. Fiber Graphite! Epoxy Lowered 1 l l 1 1 3O - Parallel Fiber olyeser Glass Lowered 1 1 l l 1 29 -- [ Nylon Lowered 1 2 l 1 1 2 -- Table 4- 2 Results from axial field measurements _Sarnple Sample Measured axial elecuic field symmetry locations for Height sample placements at the: Center Top Top Bottom Bottom Left Right Right Left Empty Cavity -- -— -- -- - - 25 27 Teflon Elevated - -- —- - - 26 raphite/ Epoxy Elevated 27 27 27 none none Perp. Fiber - Craphite/ Epoxy Elevated 27, 28 26 26 76 none Parallel Fiber Tolyester / Glass _Elevated 26 T2 26 '26 26 lon Elevated 2 26 26 _26 26 e on Lowered none Graphite7 Epoxy Lowered 25 none none none none Perp. Fiber Graphite/ Epoxy Lowered 25 25 25 _25 25 Parallel Fiber Polyester / Glass Lowered 25 none 5 '25 75 Nylon Lowered 25 25 none none none Table 4- 3 Results from axial field measurements 85 comparison or the cavi at: . Fiber Parallel Fiber =non- Table 4- 4 Frequency shift measurement results Similar field behavior as a function of sample size has also been observed theoretically by Mannering and was described as field confinement and field exclusion(Mannering, 1992). Field confinement was described as the concentration of the field in or around the sample load and conversely, field exclusion was the exclusion of the fields from the load as the sample size is varied. Field confinement was said to be the characteristic of lower order modes which require long coupling probe depths when microwave energy is coupled form the side of the cavity. This is said to be due to low field magnitude outside and away from the load and the primary axial field direction(Mannering, 1992). The field confinement theory is consistent with what is shown in this data and also serves as a verification of the generation of a lower order mode as was expected. Additionally, there is another significant finding that, although the Teflon block may be considered to be microwave transparent due to the low dielectric constant, the size of it does have an effect on the field distribution and behavior of the fields in the cavity. 86 Thus, considering the Teflon material to be transparent and non-interfering with the fields based upon dielectric prOperties is not totally correct. This can be further understood from perturbation theory, which tells us that field interactions can come about from dielectric prOperties as well as sample placement in the cavity and geometrya-larrington, 1961). In general the axial field measurements for the different sample loads at the elevated height was essentially similar to that of the Teflon loaded cavity at the elevated position as shown in Figures 4-6 through 4-9 and in Table 4- 2. As shown inTable 4- 2, for the graphite / epoxy sample with parallel fiber orientation, only one high amplitude peak was measured. The reason for this difference is not understood at this time, but is an indication of the complex nature by which the graphite I epoxy composite interacts with the field. For the lowered sample loads similar electric field measurements for the lowered Teflon block load was also observed for most of the sample loads as shown in Table 4— 2. A different measurement was observed for the graphite / epoxy load, where two high amplitude peaks were observed for the sample placement in the center and at the top right corner of the cavity. Similar peaks were also observed for the Nylon load with sample placement at the top left comer of the cavity. The difference in the axial field pattern for the center sample placement is not as surprising as that for the edge sample placements. Because these electromagnetic fields are theoretically symmetrical fields, it would be expected that similar symmetry is shown in the field measurements for the sample placement at the edges. But, this was not the case. Incidentally, it should be noted that similar results have been observed in heating experiments where the graphite epoxy composite sample show severe localized burning at the top left or tap right comers of the sample. Possible reasons could be due to the coupling probe allignment and depth which could have caused near-field interactions 87 with the sample. In summary, the axial field measurements have indicated that in general, a mode with a 2 index of 2 was generated for the elevated sample loads, and that with an index of 1 was generated for the lowered sample loads. The circumferential field measurements are shown in Figures 4-10 through 4—15 and in Table 4- 2. In these plots the y-axis is the measured E-field and the x-axis is the circumferential field measurements along five columns (3,4,5,6,7) and four rows- (24,25,26,27). There were four circumferential rows which were numbered from 25 to 27 and each row contained five columns which were numbered from 3 to 7(see Figure 43). Row 27 and 28 were at the same level as the coupling probe and column 5 is 180 degrees away from the coupling probe. Figure 4—10 shows the empty cavity circumferential field measurements for a TM(012) mode at two different probe depths at a fixed cavity length where resonance was achieved. The electric field measurements for the short probe depth at row 25 was , constant along the four columns. However, as the field measurements were taken close to the coupling probe height at row 26 to 28. the field became more variable. When the probe depth was increased to 19.28 mm the measured field also became variable for all the rows. This indicates that the coupling probe location and depth has an effect on the field pattern in the cavity. Figures 4-12 through 4- 15 and Table 4-2 show the circumferential field . measurements for the different sample loads. In these Figures the enumerated sample placements, center corresponds to position 1, t0p left to position 2, top right to position 4, bottom right to position 6, and bottom left to position 8. In Table 4-2 the third column shows the row number where constant field measurements were achieved. 88 This table shows that constant field measurements were obtained in the region around row #26 for the elevated position, and around row #25 for the lowered position for most of the sample loads. In all cases, the field measurements were more variable as the measurements were taken closer to the coupling probe. In summary, the circumferential field measurements suggest that interacu‘ons from the coupling probe due to near field could have a significant effect on the fields in the cavity. This effect could be due to the interactions between the sample and the coupling probe or the elecuic field measuring micro coax probe and the coupling probe. Ideally, the electric field diagnostic holes should be placed away from the coupling probe to avoid such interactions. As such, these results cannot be used for conclusive characterization of the loaded cavity mode. They can be used to map the field characteristics of the in-plane field pattern as a function of sample placement, which indicates that the regions of the sample that are close to the coupling probe interacts differently than those away from it. This is shown by the similarity in the field measurements for the top left and top right regions and the bottom right and bottom left legions , as shown in Figures 4-12 through 4-15. Figure 4-16 through 4-19 and table 4-3 show the frequency Shift measurement for the different sample loads. In Table 4—3, column 3 shows the frequency shift behavior along the different quadrants of the cavity. An S was used to indicate if the shift was similar to the other regions of the cavity and N for non-similar. Typically, as a sample is placed in a region of high electric field, there is a large shift in the resonant frequency shift according to perturbation theory(Harrington, 1961). As such, the frequency shift measurements can be used to map the electric field intensity in the cavity. For the polyester/ glass and Nylon samples the shifts were essentially similar in all four quadrants of the cavity. However, for the graphite I epoxy samples the freq‘lency shifts were different in the 90 ° - 180 ° and 180 ° - 270 ° quadrants of the c - . . . . . . . aVIIY. Wthh rs consistent With what was observed in the Circumferential field 89 measurements. The measurements did not suggest a strong function of fiber orientation for the graphite epoxy samples. The frequency shifts were always greater for the elevated samples placements than for the lowered sample placements. This is an indication that the field intensities are higher at the elevated sample location than at the lowered sample location. To this point the concept of near-fields interaction has been mentioned and not explained. Near field interactions are known to occur in regions close to the microwave coupling device. At regions close to an antenna, (about half of the free space wave length), it is common to observe the effects of near-fields(Risman 1987). Near-fields are conditions in a region in space where the time average of the non-radiating energy density exceeds the radiating energy density. Thus, in the presence of near-fields the electromagnetic field distribution can be significantly altered. The result of altering the field distribution is that the electromagnetic energy distribution becomes less predictable and heating uniformity becomes less predictable, if not impossible. It would seem logical to design a cavity such that the samples are placed at a distance away from the antenna to minimize the near-field effects, however it is not that simple. Not all of the near field effects are bad. It becomes a problem when the electric field patterns are altered such . that heating uniformity is lost In the cavity used the phi-axis radial electric field measurement holes were located at the same vertical height as the coupling probe which could have contributed to the near-field interactions. 4.3-3 Loaded cavity mode estimation The axial electric field measurements only provided conclusive information about the Value of the z-index of the loaded cavity mode. The circumferential field measurements were not consistent enough to confirm the presence of a (p-symmetric mode. However, the field measurement from the top of the cavity did indicate the exiStlerrce of a TM. Hence, it was assumed that the loaded cavity mode used in the 90 measurements was a TM-mode that would have been excited at an empty cavity length of 14.39 cm versus the 12.97 and 11.97 cm in the loaded cavity. It was also assumed that the decrease in the empty resonant cavity length was due to a summed effect from the Teflon block and the sample load. However, since the samples used in the previous measurements were quite small, the decrease in the empty cavity length was assumed to be due mainly to the Teflon block Using the frequency shift method discussed above, experiments were done to measure the resonant frequency shift for a 3-inch square by 0.25 inches thick Nylon, polyester/ glass, and epoxy / graphite sample loads. In these experiments the Teflon loaded cavity resonant reference of 2.45 GHz was used as the reference. The cavity length decrease was assumed to be proportional to the change in the resonant frequency and was calculated from the following: Ad = cl_7::I-f_l' f 0 Ac, - total cavity length decrease due to sample load EC, - empty cavity length, cm 7;, - Teflon loaded cavity length, cm f1 - sample loaded cavity frequency, GHz f0 - Teflon loaded resonant frequency, GHz Since it was assumed that the loaded cavity mode was a TM(012) mode, the resonant empty cavity length was 14.4 cm. Table 4-4 summarizes the results for the cavity length shift calculations for the three different sample loads. The Teflon loaded cavity lengths that were obtained from the previous measurements for the lowered and elevated sample shown in column 2. The actual center frequency for the sample loaded cavity is shown in column 3, and the total cavity length shift in column 4. The lowered sample cavity length shift for all the samples was between 1.6 cm and 1.8 cm, and that for the lowered sample placement was between 2.6 cm and 2.9 cm. Hence, it was 91 c1. 1 (cm) (GHz) Table 4. 5 Total cavity shift for a loaded cavity for a TM(012) mode assumed that the loaded cavity length shift due to the samples and Teflon block for the lowered sample placement was approximately 3.0 cm. To verify these results, two separate'experiments were done to heat 3 inch square, unidirectional 24-ply graphite / epoxy and vinyl ester I glass composites. Samples were placed on a Teflon block with four temperature probes attached to the surface of the sample as shown in Figure 4- 1 Sample Placement. The lower sample height (1.53 cm from the base of the cavity) location was used. All tunable loaded cavity modes ranging between 8.0 cm and 20.0 cm were used to heat the samples. Each sample was heated for ' thirty minutes or until a temperature gradient of 100 or more was achieved. During heating, the cavity was automatically tuned as needed and the power level was set at 100 % output which varied between 80 Watts and 90 Watts. 4.3.3.] Results and Discussions Table 3 summarizes the graphite I epoxy heating results and table 4 summarizes the vinyl ester/ glass heating results. Five tunable modes were located for the vinyl I ester / glass sample and four for the graphite epoxy composite. In the empty cavity, 92 there are six tunable modes between 8 cm and 20 cm. In these tables, column 3 lists the measured temperatures, which are ordered from the best heating site to the worst. By using the calculations discussed above an estimated cavity length was determined and is shown in column 4. By inspecting the heating patterns, the corresponding theoretical modes were determined which are listed in column 6 along with the corresponding theoretical cavity lengths in column 5. In column 7, the actual cavity length which is the difference between the heated cavity length and the theoretical cavity length are shown. These results show that in general, the loaded cavity length shift estimation method is relatively accurate for lower order modes and not for higher order modes. This is not surprising since higher order modes tend to be closer together in frequency and cavity length and multiple-mode overlapping are more prominent which makes mode identification more complex. It is important to note that these theoretical cavity modes only indicate that the heating modes have similar characteristics as the corresponding theoretical ones. Loaded cavity modes are not necessarily unique as the empty cavity modes. They typically overlap other modes and exist as hybrids of other modes(Asmussen, 1987). The purpose for identifying the loaded cavity modes is to provide an initial estimate such that mode selection and tuning dung processing can be enhanced without the need for heating experiments to characterize the modes. 93 eating Probe Heated sites Estim. Calc. Calc. Actual CL Depth empty empty mode A 1: (cm) (mm) CL CL (Ccalc. CL) - (cm) (cm) (Heat CL) “817—. 725* T3=T4>T2>T1 11.1? 11.28 TE(Oll) 3.16 11.91 .* 72311939134 14.9 14.40 TM(0123‘749 “12.44 135 T4>fi>T3>T1 15.44 15.62_'T_3—_E( 11) 31.1 ”W . 7 18.2 T4>T2>T1>T3 16.47_—'677_1 . W21 ) 3.1T T9773 20.3 T1>T3=T4=1‘"'2‘ "2‘7—2. 3 '2—"162 "TM(013) 1.88 Table 4- 6 Graphite Epoxy Heating Results Heating Probe Heated sites Estim. Calc. Calc. Actual Cavity Depth CL empty mode A = Length (mm) (cm) CL (831C CL) _ ‘ 8.67 7.67 T1>T3>T2>T4 11.67 11.283 TM(lll) 2.56 11.55 11.39 T1>Ts>i$r4 14.55 14.404 TM(012')"2.85 WWWl 16.41 16.TK)'W‘5.0 14.03 3.97—W‘1'703 hybrid hybrid ..-- 16.18 5.12_—'T3—>fi=T4>T1 19.18 Wmus) 3.89 Table 4- 7 Polyester / glass heating results (Fellows , 1995) From the heating results in Table 4- 6 and Table 4- 7 there are other noteworthy results which do not necessarily relate to loaded cavity length estimation but to the understanding of the heating characteristics of the two different composites. Longer coupling probe depths were required to tune the cavity loaded with graphite! epoxy sample than the vinyl ester / glass sample. With the longer coupling probes, the top right or the more 4.4 1 94 or the top left comers of the sample always heated preferentially. This was also observed in the frequency shift measurements from the cavity characterization section and was said to be due to'near-field interactions from the long coupling probe. For the vinyl ester / glass sample, the edge and center heating sites were more uniquely defined than for the epoxy / graphite sample. This could suggest that there is more mode overlapping with the graphite / epoxy sample. This is not surprising since the graphite I epoxy sample is a more complex system electrically than the glass vinyl ester/ system. 4.4 Summary All theoretical cavity modes available in the 8.0 cm radius by 40 cm long cavity were calculated theoretically and measured experimentally. A total of 17 modes were calculated theoretical of which 15 were located in the empty cavity using the low power sweep oscillator. There were 12 TE—modes, 5 TM-modes, and 2-degenerate modes at theoretical cavity lengths of 11.283 cm and 22.566 cm. The cavity quality factor was calculated for each mode and ranged from 4,000 to 23,000 where the TE modes have the higher values. The loaded cavity characterization results showed that : a) Although the Teflon block used for sample placement is electromagnetically transparent and would not preferentially heat, it does have an effect on the shape of the field pattern in the cavity and thus the heating pattern , as was shown form the axial electric field measurements. 1)) Sample placement height can affect the field distribution along the z-axis as was shown in the axial field measurements. c) Longer coupling probe depths have an effect on the circumferential field pattern due to near-field interactions as was shown in the circumferential field measurements. d) e) 8) 95 There was field symmetry when the samples were placed at a lower level, away from the coupling probe than at the level parallel with the coupling probe as shown from the frequency shift results. Variant field behavior was more prominent for the graphite epoxy composite than for the polyester glass or the nylon samples in all the measurements. Variant field behavior was more pronounced at the elevated sample location than at the lowered sample location in all the measurements. Variable field behavior was more pronounced at regions close to the coupling probe than away from it as was shown from the circumferential field measurements. A simple empirical method was deve10ped for calculating the loaded cavity modes for using graphite / epoxy and vinyl ester/ glass composite. The method was found to be more accurate for lower order modes than for higher order modes for both samples. CHAPTER 5 AUTOMATION AND CONTROL OF THE SINGLE-MODE RESONANT CAVITY 5.1 Introduction Virtues of microwave processing of composites in a single-mode resonant cavity have been well demonstrated as a viable alternative to thermal processing at the lab-scale(Wei, 1991; J ow, 1988 ; Fellows, 1992 ;Vogel, 1989 ) where the single— mode resonant cavity was operated as manual device that is impracticable as a process. The hardware consisted of four basic components: 1) tunable cavity with gear drives for manually adjusting the cavity length and probe depth, and micrometers for measuring the cavity length and probe depth; 2) an external circuit for processing with major components which included a microwave power source, power meters, and a dummy load to absorb reflected power; 3) an external circuit for diagnostics and for dielectric analysis with major components that includ sweep oscillator, power meters an oscillosc0pe; 4) fiber optic thermometry for invasive temperature measurements; 5) computer interface for automatic data acquisition of input power, reflected power and temperature for post-analysis, and the on/off control of a switch to direct the microwave power away or to the cavity. In processing, the microwave cavity was Operated as an Open-loop system where a seasoned Operator was the necessary link to close the control-loop. A typical processing activity included; 1) the continuous tuning of the cavity by the manual rotation of dial knobs to adjust the cavity length and probe depth to minimize the reflected power; 2) the frequent selection of new cavity modes by carefully adjusting 96 97 the cavity length and probe depth to new locations, while tuning the cavity and manually modulating the input power to achieve uniform heating; 3)temperature control by the automatic on/off control of an electronic switch to direct the microwave to or away from the cavity. Data acquisition was done as a separate activity using both analog and digital ( RS- 232) interface which required complicated program device drivers. Hence, the cavity was Operated as an independent device from the external circuit, and data acquisition unit and processing results varied from one Operator to the other and optimization was diffith at best. In diagnostics and dielectric measurement, the cavity is also manually tuned to generate a power absorption curve on an oscillosc0pe. This curve is then analysed by inspection to calculate the cavity resonant frequency shift and cavity Q-factor(see Chapter 3). The cavity resonant frequency shift and the Q-factor are both important parameters in the calculation of the dielectric constant and loss factor. Although, dielectric analysis the single-mode resonant cavity is considered to be one of the most accurate methods(Jow , 1989) the manual measurements technique results in inconsistent results. Thus, in order to realize the potential of this technology as an alternative to thermal processing it was advanced as an integrated process through automation. In the automation, a control system was designed and built in addition to the develOpment and implementation of a set of sophisticated and comprehensive control software programs for controlling the curing process in the cavity. The single-mode resonant microwave process is governed by discrete electromagnetic modes and a complex non-linear interactions of electromagnetics and material. Mathematical models that completely and accurately describe the dynamics of the microwave process were found to be complex and computationally intensive for conuol purposes. Although the control of the microwave process is a novel concept, the difficulty in modeling the dynamics of the process is a control 98 problem which fits a class of control topics under non-traditional control methodologies or intelligent control system(see Chapter 2). Using elements of traditional and non-traditional control methodologies, a closedIOOp feedback control system was developed to satisfy the control Objectives of; efficient energy coupling or mode tuning, mode selection and uniform heating, and controlled heating. Efficient energy coupling was treated as a mathematical minimization problem in which a 2-dirnensional simplex minimization search technique was used to implement the controller. Uniform heating was based upon heuristics and empirical data that was derived from the cavity characterization results, and controlled heating was based upon traditional PID(prOportional-integra1- derivative) method. Additionally, although the integral part of this work was focused on the automation of the processing system, elements of the diagnostics system was also automated using GP[B(General Purpose Interface Board) and AID interface. Thus, allowing for the automatic and more accurate measurement of the cavity parameters for diagnostics, and the potential for eliminating the oscilloscpe and the manual measuring methods. This chapter is dedicated to the presentation of the control software developments to automate the single-mode resonant cavity for processing. Hardware automation components and data acquisition interface are summarized in this chapter with additional information in Appendix A. The developed LabView control software program for automating and for controlling the microwave curing process is also presented. Finally, elements of the hardware and software for the automation of the diagnostics system is also discussed. 99 5.2 Cavity Automation 5.2.1 Cavity Description The designed control system included an automated sin gle-mode resonant cavity and a microwave circuit in which all the instruments were interfaced with a data acquisition unit. The manually operated microwave circuit design was essentially modified to accommodate automatic data acquisition, and the microwave power source or magnetron was modified for automatic analog manipulation. The core of the hardware automation was the design and fabrication of the cavity with mechanized drives. The cavity is a cylindrical brass tube with moving parts that include internal transverse shorting plates, and a coupling probe. The transverse shorting plates are adjustable to vary the volume of the cavity. The material load rests on the bottom shorting plate which is fixed in place during processing and is removable for material load. Microwave shielding material or gasket made out of silver called Finger Stock (V arian CF-300) are soft soldered around both base plate and shorting plate. The purpose of this is to provide good electrical contact between these transverse plates and the cavity wall and eliminate microwave leakage. A semi-rigid 50-ohm impedance brass coaxial probe serves as a field excitation or coupling probe to couple microwave power into the cavity. The probe is 0.25 inch in diameter and 2 inches long. There are a series of axial and circumferential holes equally spaced 0.92 cm apart through the cavity wall. These holes are used to measure the square of the electric field strength along the cavity walls. A 2 mm COpper micro coaxial probe is used as an E—field diagnostic probe. Figure 5-1 shows a picture of a typical cavity that is disassembled into the major components. From clockwise in Figure 5-1, there is the shorting plate which is attached to the gear driven drives, the body of the cavity, the base plate with the attached fingerstock, and the coupling probe. 100 5.2.2 Automated Cavity and Mechanized Drives The overall cavity design was not novel and had been previously deve10ped with either an axial or a radial mounted coupling probe(Asmussen, 1988). What was novel about this cavity design was the mechanized drives and the implementation of dual mounted coupling probes at the axial and radial positions. The automated new cavity was designed with mechanized drives for the shorting- plate, base plate and coupling probe. The new cavity is 18.9 cm (7 inches) in diameter and 42 cm tall which is 1.5 times taller than the manually operated cavity(see Figures 5-2 and 5-3). This increase in cavity height allowed for the excitation of 12 additiOnal modes (see table 41). Figure 5-2 (a,b,c) show a schematic of the manually driven cavity, a picture of an actual cavity, and the gear - driven manual drive, respectively. Figure 5-3(a,b,c) show a schematic of the automated cavity, a picture of the automated cavity and a mechanized drive, respectively. The mechanized drives use a belt and pulley system in which a stepper-motor drives a ball screw(see Figure 5-3 0). By driving the ball screw rather than the knot the frictional losses are minimized. There are two stainless steel 114 inch. diameter guide rods which ride in graphite bearings. The use of the linear bearings rather than the spur gears eliminated the binding problems that existed in the manual cavity. Detail discussion of the drives are presented in Appendix A. Linear motion potentiometers were installed on the cavity for measuring the cavity length and probe depth. These potentiometers were calibrated by generating voltage and length relationships for the cavity length and probe depth(see Appendix A) for details. 101 Figure 5- l Single-mode Resonant Cavity Components . From clockwise, shorting plate and drive, cavity body, base plate with fingerstock, coupling probe 102 Coupling | I l I Probe {Shorting Plate I 5% , l L211? 'Lc 1 Figure 5- 2a Schematic of manually Operated single-mode resonant Cavity 103 Figure 5-2b Picture of manually operated single-mode resonant Cavity Figure 5-2c Drive for manually operated single-mode resonant Cavity 105 II - Short-plate V/Il/lm Will/IA Drive . hort-plate I inger Stock Top Coupling Probe Sargle 1:: '— l 1 Side Coulp 'ng ~ - Probe Figure 5- 3a Schematic of automated single-mode resonant cavity 106 Figure 5-3b Picture of automated Single-mode resonant cavity 107 Figure 5-3c Mechanized drive for automated single-mode resonant cavity 108 Stepper Motors and Drivers 5.2.2.1 Stepper Motor and Driver Hardware The stepper motor is an 8-lead motor and driver unit which was designed specifically as a high performance positioning device(see Frgure 5-4). The structure was constructed with class B insulation material which is capable of withstanding temperatures at the motor coil of 130°C(255 F) with no reduction in motor life. It has a holding torque capability of 118 oz-inch and variable step angles of 0.90 and 0.450 which corresponds to 400 pulses per revolution and 800 pulse per revolution, respectively. Table 5- 1 summarizes the corresponding cavity length and probe depth adjustment speeds. Movement of the stepper motor was accomplished by regulating the number and frequency of pulses sent to the driver's terminals. The motor rotates one step for each pulse received at the pulse terminal and the direction of rotation is controlled by the signal applied to the driver's "CW/CCW" terminal. The interface to the drivers was by a non-standard TTLCI‘ransducer Transducer Logic) digital logic IC(integrated circuit). A signal at (4-5 V) was defined to be high while a signal at (0-0.5V) was Motor Pulses Speed of Cavity length Speed of Probe depth Pulses I revolution adjustment adjustment [cm/min] [in/min] [cm/min] [in/min] 800 10 25 lST 70 400 5 12 90 35 Table 5- l Cavity Length and Probe Depth Adjustment Rates 109 W _ {-13 + ”@— lPulse GB + SW1 —€§_ c ccw 1'1 ‘E —— - 1-2 —= SW1 1.3 =- 2-1 1'4 "=1 ‘ Stri d—— ~ S 1-5 1::- fl pe "7 2.- . . 8 o O 2 g ('9 Red 2 : FE- Orange ,_. Yellow ”g lACllS ‘ {-13 PG 1 l Striped Stepping Black Motor Red Orange Yellow Figure 5- 4 Stepper Motor and Driver 1 10 said to be low. In a standard TTL interface a signal at (2.4 V- 5.2V) is defined as high and that at 0.0V-0.8 V is low. Thus, a standard TTL wiring could not be used and a 5V external power supplyCI‘ucker Power Supply) was necessary to drive these tenninals(see Figure A-2 and Appendix A). 5.2.2.2 Stepper Motor Driver Software The stepper motors were driven by sending pulse signals to the driver. The pulse signals sent to the stepper motor drivers were generated as a sub-program called “Move Read.vi” using the LabView software(Appendix E). The input to the software driver were motor direction and number of pulses. Given this input a DO- loop is iterated N times where N is the number of pulses. For every iteration a high or low voltage signal is sent to the pulse terminal to move the motor. This pulse is generated by comparing the remainder(R) of the iteration (i) number divided by two, to one. If the remainder is equal to one then high value is output and a low value is output otherwise. The output voltage is alternated between high and low to generate a continuous pulse to the driver pulse terminal. A logic diagram is shown in Figure 5-5 and its equivalent in LabView is shown in Figure 5-6. In the LabView program the DIG-LINE box represents a subroutine for driving the digital line on the data acquisition board. lll Ouput = 0 (Low) or 0 ~ 0.5 V V k J Figure 5- 5 Program Logic for Stepper Motor Driver f N glumber of Pulses I are [Pulse Terminal B B11" lu- Eli." Ed 10 k J Figure 5- 6 LabView Program Version of Figure 5-6 1 12 5.2.3 Cavity length and Probe Depth Measurement The cavity length and probe depth are measured by a 10000 ohm linear motion potentiometers with stroke length of (X and Y), respectively. The potentiometers require a 5V reference. The Operation of the linear motion potentiometers can be described by the principles of variable resistors. It consist of a resistance element, and a sliding arm that makes contact with the stationary resistance element, see Figure 5- 7. A Tucker variable power supply is used to supply the 5V reference required by the potentiometers. There is a local digital display for the potentiometer readings. This indicator requires a 9V power supply which is also provided by the Tucker variable power supply. The potentiometers are calibrated by determining the relationship between the cavity length and probe positions and output voltage. A linear relationship was determined for the voltage and cavity length and probe depth positions as in Equation (51) Cl: ax(V)+kc Pd = p x (V)+ kp (5'1) where C1 = cavity length [cm] Pd = probe depth [mm] or = slope [Volts/cm) B = slope [Volts/mm] kc = y-intercept [cm] kp = y-intercept [mm] 113 B Slidthr-In I] 3 5,, WV“ :lc C o 1th L J Figure 5. 7 Linear Motion Potentiometers 5.3 Microwave System Automation 5.3.1.1 Automation of External Circuit The processing extemal microwave circuit is made of the following components, magnetron microwave power source, power meters, dummy load, circulators, directional couplers and a crystal detector see Figure 5. 8 and Figure 5—9. The diagnostic system consist of the same components except for that the power source is replaced by a sweep oscillator and the power meters are replaced by the oscilloscope. Coax cables are used for all connections and all devices are rated for a power level range of 0100 Watts at a frequency of 2.45 GHz. The magnetron is a continuous wave, single frequency, multi-power microwave power source which Operates at 2.45 GHz and a power range of 0-100W. Incident power from the magnetron is decoupled by a directional coupler, where one end is directly sent to the cavity and the other end is attenuated and sent to a power meter. A circulator directs the incident wave to the cavity and the reflected 114 Ewe: Saw a... a. I .8: 1 1 t 3::th we: on: a 555.4 - . _< .8555 _ I a .355: I w J at... ._ .I. 58058.58.— ..Bm >3 no. a. , , 2.8m 9:...an , . .a:e:uo..5 I a 32.5 L»:— , , I. I III «33:5 3.2—83:85— I .5 a :9. - 8a:— uEtesm I -A a _ .0 09:8 2 , ,, I .83... a .38:— :836 [+2an one...— u==a=eU_ 33:23: ueflafimlu 3.2-.80 . baa—am .5.— ._ .53.... d .88.: I - - seem—ELI] - 3&8.“ 8a..— owe: .52.... d .88:— - [1* .2596 82m 95.8.5 . Tacoma—=86— r L . - 33:0 .2.:-.- .... . . ‘u I... _ f .a .. I Figure 5- 8 Schematic of External Circuit 115 Figure 5- 9 Picture of Automated Microwave Processing System 1 l6 wave to a dummy load. The temperature sensing system uses fluoroptic therrnometry technology and is manufactured by Luxtron(Luxtron, 1989). The system consist of a sensing unit and fiber optic probes which have a capability for measuring temperature in the range of -190 to 300 0C. It is made up of a silica fiber with a Teflon jacket body and sensing tip made up of magnesium fluorogerrnanate. The sensing unit is made up of Optical heads, Xenon flash lamp, lenses and a filter. To sense temperature the Xenon lamp emits a blue light which excites the sensor and causes the sensor to exhibit a deep red fluorescence. The time at which it takes the intensity of the red light to decay is exponential with time and is inversely related to temperature. Each unit has 4wchannel fiber Optic temperature sensing probes with a capability for local display, analog output and digital output. ‘ In the microwave processing circuit, the power meters and temperature measurement were interfaced using analog output (AI). This was an improvement from the previous state of technology where the temperature measurement was interfaced using digital input (RS-232) and the other meters were interfaced using analog input. Although, this was a perfectly valid data acquisition system it required very complex device level driver programming which was very difficult to maintain or to add additional sensors. The power meters and the temperature sensing units are analog sensing devices with a voltage signal range of 0-5V for a zero to full scale values. These analog devices were available with analog output connections as BNC type connectors. A Coax cable with one end attached to a BNC type connector and the other end stripped as the conductive wire was used to connect the analog devices to the data acquisition board. The connection to the data acquisition board was done using “single-ended" connection configuration, where one end of the wire is hooked to the channel and the other to ground on the terminal box. 1 17 5.3.1.2 Automation of Microwave Power Source The magnetron, microwave power source was interfaced as an analog output to the data acquisition board. The source was augmented using integrated circuitry to provide for the analog output interface. This was a major contribution to the external circuit which allowed for the automatic regulation of the microwave input power. In the previous state of technology, the source was not directly controlled but the output was either directed to or away from the cavity using a digital switch. The power directed away from the cavity was directed to a dummy load which was wasted which was an inefficient use of the microwave power. Additionally, the onIoff or band-bang control method did not provide for a wide variety of processing flexibility. The magnetron modification allowed for both manual and automatic regulation of the output power. See Appendix A for details of operation. The voltage range of 0-100 W of the source corresponded to an analog output range of 0-1.112 V in an almost linear relationship Figure 5- 10. Magnetron Input Power vs. Voltage Calibration Input Volts(V) 0 5 t : 4 0 20 40 60 80 100 Output Power (W) Figure 5- 10 Magnetron Power versus Voltage Calibration for Analog Control 1 18 5.4 Data Acquisition 5.4.1 Data Acquisition Hardware The manual system external circuit was essentially automated by interfacing with a data acquisition board. A National Instrument data acquisition and control board NBMIOl6H was used. This board is a high performance multi-functional analog, digital, timing input/output board for the Macintosh computer. This board has a maximum sampling rate of 47 K samples per second. It is a 12-bit successive approximation ADC(analog to digital converter) with 16 analog inputs, two l2-bit DACs with voltage outputs, 8 lines of TTL-compatible digital IIO, three 16-bit counter/timer channel for timing I/O, 2-(5 V) power supply, l-digital ground, 1- analog input ground, and l-analog output ground(see Appendix A for details and Table A-2). 5.4.2 Data Acquisition Software Program Data acquisition is one of the main attributes of the LabView software package. It is provided with several data acquisition device drivers (subroutines) which are easily modified to meet any data acquisition needs. These drivers are available in the data acquisition library. Typically a configuration subroutine is required to configure all the device channels where data will be acquired, coupled with the data acquisition subroutine. The data acquisition subroutine is used to set the signal type and signal ranges and the number of data points to acquire per channel. The subroutine “AI Group Config” was used to configure the channels and “Single scan” was used to acquire the data values (see Appendix E). In this work there were 16-temperature channels, 2-potentiometers channels for cavity length and probe depth, and 2- power meters for input and reflected power. The temperature, power meters and potentiometers were interfaced as analog input. The stepper motors were interfaced as digital output, and the power source was l 19 interfaced as analog output. To minimize noise in the acquired data, each channel was scanned multiple times (10 times) and the values averaged. Scanning rate was set at 1000 samples per second. Up to this point the elements of the system hardware including the cavity, data acquisition and its interface with the external circuit have been discussed. A schematic of the overall microwave processing circuit is shown in Figure 5-8 and a picture in Figure 5-9. In the sections that follow the control software for the controllers in the controlloop are discussed in detail. These components are the mode tuning controller, mode selection controller and the power controller as shown in the controlloop in Figure 1-1 and in this chapter as in Figure 5-11. 120 It. 882 .2595 .202 A 53a 82.. _ i 852 Samoa _ 8an 8.8mmw. _ .3. 53.... 8.8.3.. r. Sweet. 5.2.0 _l_ I111 3.65.50 9.2.... 682 .8521. Sodom 0523...»... L 1. L 85$. 23% 58: 88.. 555. .395 , - 598.. 3.2.0 - 5.2.0 .5551. _ _8m5m 053.38%. . 5... 2282.20... 85:32 I .O..o=._eu 53.... 258.8. .6... 2.28.0 2388...»... 8.85.50 5.82% 0.52 9:898. ...< 0:698. .r Figure 5- 11 Control loop from Chapter 1 121 5.5 Process Control Software 5.5.1 Efficient Coupling— Mode Tuning For a given cavity geometry, dimension and microwave excitation frequency, the theoretical cavity lengths at which different modes can be excited can be calculated(mode chart). This calculation is done from the cut-off frequency Equations (3-16 & 3-17) which are derived from the eigenvalues to the solution of Maxwell's equations. Once the cavity is loaded with material the empty cavity modes become modified(Asmussen, 1974) and sometimes become in general hybrid modes(Harrington, 1961 ). Modification of the theoretical modes is a function Of material placement, volume, shape and electrical properties. The dependence on electrical properties introduces additional material conductance and susceptance to changes in the external circuit. As the material is heated the dielectric loss factor changes, along with material geometry and the required frequency of the microwave source to maintain the mode. Since the changes that alter the resonant frequency are not available, due to unavailable sensing techniques or prohibitive computation, it is difficult to develOp a control system for tuning from traditional control methodologies. Hence, a radical approach is used where mode tuning is described as a mathematical function minimization problem. 5.5.1.1 Approach This approach is based on the justification that, conceptually mode tuning is a minimization technique which can be described by a 3-dimensional surface. Mathematically, mode tuning can be described as the minimization of a convex objeCtive function with constraints. For this type of problem an analytical search 122 technique could be used if the Objective function is readily continuous and differentiable and could be expressed as a mathematical function. This function could be either as a result of mathematical modeling or curve fitting of a numerical data(Beveridge, 1970 ). Stated differently, the analytical technique would require extensive knowledge about the objective function, but will provide all to the function, which is nm the case with the numerical methods. In mode tuning, extensive knowledge about the objective function cannot be easily derived apriori, and all minimums need not be found. Thus, a numerical search technique was chosen to be a sufficient method. The general principle of the numerical search methods can be described as what is called ”homing in"(Boas, 1963 ). These search techniques are based on the determination of a base point, from which a search method selects a new set of independent variables and tests the objective function to see which set gives a better value for the objective function. Based upon the comparison, another set of variables are chosen and the objective function re-tested and re-evaluated. The key is to Optimize the set of variables selected such that wasteful computations and experimentation are minimized. There are different numerical methods that can be classified as direct search methods or gradient search methods. Direct search methods require only the evaluation of the objective function at a particular location, while the gradient search method requires both the evaluation of the objective function and the gradient. The direct search method may therefore be found to be efficient as far as computation time, although it might not move in the best direction every time. Instead of seeking directional accuracy in movement, the emphasis is placed upon speed of computation and movement. 123 5.5.1.2 Simplex Method The simplest direct search method is the univariate method. In this method where the tOtal number of variables is 11, one variable is changed at a time by keeping n-l of the 11 variables fixed. This technique has the tendency to oscillate as the optimum is approached and may not even converge, depending on the shape of the minimum, i.e. if the minimum is narrow as compared with the step size, or if the minimum is not parallel to the coordinate axis(Beveridge, 1970 ). Thus, the univariate method is not robust and may be applicable to very specific cases where the variables do not interact and each variable can be optimized independent of the other. However, in actual situations variables do interact, and a more robust approach would be not to fix the direction or the step size of the search direction, but to permit these parameters to change as a result of experimental data or evaluation of the objective function. A good illustration of such a method would be the Sequential Simplex Method which was first prOposed by(Spendley, 1962 ). It is a highly efficient, multi-factor, empirical feedback strategy that requires neither the large number of experiments nor complex calculations of the evolutionary Operation(Morgan, 1974 ). A modification by Campey and Nichols(Campey, 1961 ), Nelder and Mead(Nelder, 1963 ), and Box(Box, 1965 ) of the original simplex method provides the capability for acceleration in directions that are favorable and deceleration in directions that are unfavorable. These simplex systems are conceptually different from the simplex methods used in linear programming(Beveridge, 1970). The method takes a geometric figure, known as a simplex as a basis. A simplex is a geometric figure defined by a number of points equal to one more than the number of dimensions of space. A simplex of two dimensions is a triangle, a simplex in three dimensions is a tetrahedron. Experiments or evaluations are 124 organized such that the objective function is evaluated at the points formed by the vertices of the geometric figure. One vertex is then rejected as being inferior in value to the others. The general direction of the search may then be taken in a direction away from the center of gravity of the remaining, being chosen so that the movement passes through the center of gravity of the remaining points. A new point is then selected along this direction and the search proceeds by the process of vertex rejection, regeneration and reflection, and figure expansion until the figure straddles the optimum and is contracted to the optimum. The simplex method adapts itself to local landscape, elongating down long inclined planes, changing direction on encountering a valley at an angle, and contracting in the neighborhood of the minimum. It is simple to program, and has the advantage of needing only the smallest number of points to start and only one new evaluation for each movement. Its main disadvantage lies in the necessity to scale the problem beforehand so that unit changes in each variable are of equal interest in its ability to accelerate. This is the method that was programmed as the tuning rule in the single-mode resonant cavity. 5 - S .\ .3 Stmmexmgtc mime - \j\iOt the inaction value at tpo‘mth Vt. -hht‘tgm as the shifts such that \j \\ = max hm - \ tlow} as the suffix such that y\ = min KY '1) - n (in between) as the suffix such that yh < y“ < y\ 'Pbar as the centroid of the points with i not equal to h 125 An initial base simplex is defined by a set of coordinate axes with each representing a Best, Next—to-Best, and Worst point, respectively see figure 52. At each stage of the search the worst point, P1, is replaced by a new point using the Operations called, reflection, contraction, and expansion. Reflection The coordinate of the reflection point is defined as: P‘ =(1+a)F—aP,,, (51) where P" is the reflection point and a is the reflection coefficient which is a positive constant that is greater than one. If the reflection point generates a function value y*t which is between yh and y], then Ph is replaced by P* and the a new simplex is formed. Expansion If y* < y] , i.e. if the reflection point is better than the best point, then the current reflection point is expanded again as: . P“ = yP‘ + (1+ y)75 (5-2) where gamma is the expansion coefficient, and is greater than unity. If y"< y1, replace Ph by P” and restart the simplex operation, but If y” > y] then the expansion has failed so replace Ph by P* and restart. A failed expansion may be thought of moving passed the minimum. 126 Contraction If on reflecting P to P*, y* > y1 for all i not equal to h, i.e. that replacing P by P*, y* becomes the worst point. Then a new P1, is defined to be either the old one or P* whichever has the lower y value, and form: P“ = 91’. + (1 -fi)1"> . (54) where beta is the contraction coefficient and lies between 0 and 1. P}. is then replaced by P" and then the simplex is restarted. If the contracted point is worse than the better of P" and P}, then the contraction has failed and all points are replaced by: (Pi+Pl)/2 and the simplex is restarted. Failed contraction can occur when a valley is curved and one point of the simplex is much farther from the valley bottom than the others, and thus contraction may cause the reflected point to move away from the valley bottom rather than towards it. A flow diagram of this logic is shown in Figure 5-11, implementation in LabView is in Appendix E. To further illustrate this method an example is used below. Consider three base points identified as B for best, N for next-to-best, and W for worse which form the vertex of the initial simplex BNW in Figure 5-10. P is the centroid of the face remaining when the worst vertex is eliminated. 127 Figure 5- 12 Simplex Diagram Reflection is accomplished by extending the line segment WPbar beyond Pbar to generate the new vertex R. If the response at R is more desirable than the best point B, then an expansion is attempted to E. If the response at E is not better than the best point B then expansion has failed and BNR is the new simplex and the Operation is restarted. If the response at R is between B and N then neither expansion nor contraction is recommended and BNR is taken as the new simplex. If the response at R is less desirable than the response at N, then a step has been taken in the wrong direction and the simplex is contracted by two possible methods 128 1. If the response at R is worse than the response at N but not worse than that at W, the new vertex should lie closer to R than W and the new simplex becomes BNC;- 2. If the response at R is worse than the previous worst vertex W, then the new vertex should lie closer to W than to R and the new simplex is BNCw. A failed contraction is when the result at C; is worst than the result at R or the resultath is worse than the result at W. In a situation like this further contractions are recommended(Nelder, 1963 ). The simplex method described above was used in the on-line cavity tuning to maintain the cavity at resonance. The method described is for functions without constraints, however for the tuning prOblem there are constraints on both dependent variables, which are the cavity length and probe depth. Thus, this method was augmented to compensate for these constraints. For example, when a value was calculated that was out of bounds, the value was set to the boundary value of that variable. The termination was determined from the comparison of the reflected power to a set minimum value. 129 ( >Generate Base Points j Calculate Centroid, P P=(B+N)I2 Calculate Retection Point, R I ' Evalaute function at R to be y* no * * es lisy*yi, y isy*>yh? no i neq h? es + y y es Replace Ph y P“ Ca culate y" no for expansion l is y}, < yl ? 12° . Calculate yi" 7 for contraction y... . t . e, I is y” >yh ‘2 ** Replace Pb by P v “0 Replace all Pi’s Replace Ph by P“ by (Pi-WW2 Replace Ph by P" .__no +- has rru'nimum been reached? |-—> yes +Exit Figure 5— 13 Simplex Logic Flow Sheet The application of the simplex search method to tuning the cavity in 2- dimensions is unique. This simplex method has been used in the chemical industry(Baasel, 1965 ;Carpenter, 1965 ;Kenworthy, 1967 ;Lowe, 1964 ;Lowe, 1967 ). Emst(Emst, 1968 ) was the first to use simplex designs in analytical J 130 chemistry by varying linear and quadratic y-axis gradient controls to rapidly Optimize NMR magnetic field homogeneity. It has also been used in designs in analytical chemistry(Morgan, 1974 ). Other methods such as the gradient method have been implemented for tuning the cavity in l-dimension. Alliouat(Alliouat, 1990 )successfully used it in the automatic tuning of a single mode resonant cavity for the sintering of ceramics. In this system, only the cavity length was tuned as function of the reflected power and variation in the microwave source frequency. Remember that the gradient method requires the evaluation of both the function and gradient values at any position, with the gradient being estimated by local exploration, using several experiments. Therefore, the successful implementation of this method in a l-dimensional tuning does not necessarily guarantee the same result in 2-dimensional tuning problem. 5.5.2 Mode Selection and Uniform Heating 5.5.2.1 Approach In a single mode resonant cavity, there is a unique cavity length, probe depth and field pattern associated with each mode. The theoretical cavity length at which a mode can be excited for a given cavity radius, can be calculated from the cut-off frequency equations which were discussed in Chapter 3. The theoretical electric field patterns associated with these modes are also derived from the solutions to Maxwell's equations as discussed and plotted in Chapter 3, where the field patterns indicate the electric field intensities in the cavity. Examination of these field patterns indicate that the electric field intensities are typically concentrated in the center or along the edge in a plane. As a result, a single mode alone may not be sufficient to uniformly heat a sample. This is not a unique characteristic of the applicator, but of the nature of electromagnetics. In l3 1 multi-mode ovens, i.e. home microwave ovens, this problem is addressed by the random introduction of several modes using a mode stirrer(l-luack, 1969 ). Other approaches have been the random sweeping of frequencies to introduce different modes at fixed cavity dimensionsCWei et. al., 1994). Fundamentally, the basis of all of these methods is the introduction of a variety of modes using different strategies. Some of the more pragmatic methods for achieving uniform heating in the different applicators have been presented(Adegbite et. al., 1995). One of the methods for achieving uniform heating is dual coupling which is presented in Appendix B. Another method presented called mode switching is used as the method for achieveing uniform heating in this work. Mode switching has been used to uniformly heat polyimide panels (Fellows, 1993), heat and cure complex shaped and planar vinyl ester glass panels(Fellows, 1994, Fellows, 1994). A detailed discussion of this method has been presented by (Fellows, 1994) and will be summarized here for continuity. Mode switching is a feedback mode selection strategy which utilizes a pre-deterrnined heating temperature profiles to select optimum electromagnetic modes during processing. The objective of this method is the identification of complementary temperature heating profiles which can be superimposed on each other to achieve uniform heating. Conceptually, consider the combunation of the two theoretical modes TE(011) and TM(012)(see Figure 3-2 and Figure 3-3 for elecuic field patterns corresponding to these modes) to uniformly heat a surface. For the TE(011) mode alone all the electric field intensity is concentrated at the edge of the sample in the form of a ring, and for the TM(012) mode it is concentrated in the central region. 5.5.2.2 Mode Selection Logic The mode switching implementation in this work differs from the rigorous approach presented by Fellows(Fellows, 1994). In this method all the modes in the 132 cavity were characterized from the theoretical field patterns to be either center or edge heating(see Chapter 4). Thus, to achieve uniform heating, modes were switched between groups of center heating modes and edge heating modes. The success of this method hinges upon the accuracy of the loaded cavity mode characterization into the groups of edge heating and center heating modes. As was discussed in Chapter 4 sample placement, size, and electrical properties can affect the field pattern in a complex manner that cannot be readily calculated or measured. As such, theoretical modes and elecuic field patterns cannot be linearly extrapolated to loaded cavity ones. An empirical method was used to characterize the cavity by determining the relationship of different sample and sample placement on a resonant mode. Results of these experiments were used to estimate loaded cavity lengths and corresponding elecuic field patterns. Once the loaded modes were identified, the electric field patterns were then identified and grouped into center and edge heating modes which were used in the mode switching method. Details of this work can be found in Chapter 4. The program logic used for mode selection is shown in Figure 5- 14. Implementation of this logic is depended upon the ability to group the loaded cavity modes into center and edge heating ones. These edge and center heating modes are required input on the LabView program front panel. The LabView representation of the mode selection logic is shown in Level 3 of the main program in Figure 5-18 through Figure 5-20. In this mode selection logic, the temperature gradient site must be classified into the central or edge regions of the part. This directs the program to look for a mode that would heat the cold region. To illustrate this logic a a cold center scenario will be used. Once the cold site is known to be the center it is determined whether the same modes had been succcessful in heating the center region in “Same Center Mode?”. If it is yes, then one of the previuosly classified edge heating modes is tried because it is assumed that the selected and estiamted 133 center modes are not successful. If no, then a new center mode is selected sequentially from the center heating modes list and the list is updated by adding new center heating modes if necessary. It is then determined if the new mode is the same mode that is currently being used in “Mode =Previous”. If it is yes then another mode is tried and the iteration continues. r Select Center Mode from List _ Update List Yes ame Cente u ode 9 ode = Previous No Select Center Mode from List .. Update List New Mode Figure 5- l4 Mode Selection Program Logic 134 Note that before the mode selection logic is activated, the cold region has to be cold for a period of time. This is a critical parmeter that must be determined from rigorous mass transfer models of the curing composite. Currently, this is a variable on the front panel called “Mode adjustment time”. It is typically set to 15 seconds or 5 iterations through the loop. This value was deten'mined from calculations of the time constant of a typical curing system which was found to be 3 seconds. Thus, theoretically this should allow for a 5°C temperature differential before selecting a new mode. 5.5.3 Temperature Control 5.5.3.1 Approach The approach taken in this work is the analog output regulation of the microwave power source using the measured temperature of the sample. This is different from what is typically done where the power is turned off and on in a fixed cycle. Jow (J ow, 1988 ) developed a temperature control scheme for processing composite materials where the microwave power source was turned off by a switch as the temperature setpoint was reached. In this system a conventional PID(proportional-integral-derivative) controller is used to regulate the input power from measured temperature values. Since more than one temperature value was measured, the maximum temperature value was used as the control element. 5.5.3.2 PID Method The PID routine used was available as a package in LabView. It uses an interacting positional PID algorithm as discussed by Shinskey(Shinskey, 1988 ). It supports anti-reset windup and the derivative action is done on the process variable 135 rather than on the error. This was done to stabilize the response of the derivative control action. The differential form of the PID equation: At UN D ml c[el I ;e A! (C! t—Al )] was expressed in a preferred way for implementation in three parts as: Derivative action + At(c — y) (5-5) At+DlKD where y is the output of the derivative filter which has a time, DIKD, and K1) is the derivative gain limit, D is the derivative rate in minutes per repeat. The equation is arranged such that setting the time constant to zero will produce y=c. Proportional action m = biKc[r-c-Kd(c-y)] m > mmm = mh (5-6) m < m,,m = m, where m is the manipulated variable and is limited to the high and low limits of the manipulated variable, and b is the bias which is the output when the error is zero. The bias could be fixed or could be the result of feedback from the output of the controller to produce integral action. Integral action At(m-b) +__._... At-l-I b = b (5-7) where I is the integral reset in minutes. 136 The setpoint, process variable, and output are expressed in percent. The controller was tuned to determine the proportional gain, derivative gain, and integral gain for the magnetron microwave source. The magnetron was also calibrated to determine the input voltage and output power relationship in Figure 5- 10. 5.6 LabView Curing Process Control Software Program In the previous sections, details of the specific controllers that make up the major components of the controlloop were discussed. In this section the integration of these controllers into the curing control program as implemented in LabView is discussed. In LabView the program was deve10ped as if it is operational on a device with a front panel where all input and output interactions occur and a back panel where the program is stored. 5.6.1 Front Panel The front panel for the curing program includes input and output information for using the program. Most of the information on these panels are self explanatory and can be readily understood at any level, which is one of the attractive attributes of LabView. The discussions that follow will only show how some of these front panel icons are used in the program. Due to the graphical nature of the programming code, an acceptable print quality output cannot be generated for the dissertaion and the program will only be discussed. On the input panel there are options for data storage time, power control tuning parameters, scaling values for the analog devices and data acquisition parameters which are all hardware inputs that may not require new inputs every time that the program is used. There are also input parameters for the “Mode Tuning” program which were defined in the mode tuning software program section in the previous sections. The required inputs that are sample dependent are “Glass mode 137 parameters” and the “Graphite mode parameters”. These are interval setting for loaded cavity modes. They essentially represent the regions to search for a mode for the cavity length and probe depth which are used by the mode tuning program. There is also another input for cavity length and probe depth corresponding to each mode which are input into “Cl Graphite Modes” and “PD Graphite Modes”. Once these cavity length and probe depth values are input to the program, they are indexed sequentially so that through the program they are referred by number. Another required input is the classification of modes into “Center heating” or “Edge heating” by the indexed mode number. These inputs are required for the mode selection program as an initial guess values. Finally, there is the option for the “Matrix IFiber” which automatically sets the cure temperature setpoint. Currently, there are inputs for epoxy/ graphite and polyester/ glass. It is important to note that all these inputs have default settings which can be used if other input values are not known. The output or display panel is probably one of the strengths of the LabView software program. It is very easy to generate and yet very useful in process monitoring. There is a spreadsheet at the top of the page for displaying the absolute values of all the sensed parameters. Additionally, there is a temperature strip chart and a temperature color gradient map, “Temp Probe Site”. This color map indicates the local temperature at the different sites of the sample with colors ranging from green to black, which corresponds to room temperature of 5 °C above set point. There is also a theoretical electric field pattern indicator which shows the estimated electric field patterns based upon the cavity length in the form 3-D and density plots. These plots were imported from a Mathematica file as was discussed in Chapter 3. Another important feature of the display panel is the on-line interaction with the operator. All values on the front panel that are inputs can be modified while the program is running. Indicators are always shown with arrows on the side of the 138 display box, such as “T grad Spt” located under the “Temp Probe Site”. One very important feature of this front panel is the “Shutdown” toggle switch. This switch can be activated by using the mouse to click on it. It can stop the program at any level of execution. It is a safety device which provides an additional level of flexibility for operator interaction. 5.6.2 Program Implementation The program was developed using a graphical interface programming method(LabView,l994). In this programming method the program codes are built into icons or virtual instruments (V1) with inputs and outputs which are accessed by lines or wires. A flow diagram of the microwave process control program logic shown in Figure 1-2. In LabView the overall logic is implemented using sequence structure which provides for the sequential ordering of the program logic. In the first level of the program sequence data acquisition, data scaling and data formatting for storage and display are done. In this sequence the “Cl/PD Pwr Scale” V1 is used which is a subroutine written specifically for scaling the cavity 139 K—J Acquire and Scale Da @ Yes Power Controller 1 No I Yes Mode Selection Controller No C I 7 No J— , Controller l w Mode Tuning | / length, probe depth, and power reading from voltage to actual values. (see Appendix Figure 5- 15 Program Flow Logic (also in Chapter 1) E for details of program). Other elements of the program use standard tools which basic knowledge of LabView is required to understand. Level 1 of the sequence structure is for implementing the power controller. The main VI used in this sequence is the “Pwr cntrl” which contains the PID logic discussed in the previous sections. Output of this V1 is a voltage value which is sent to the power source via analog output. Level 2 implements the “Mode tuning controller” by calling other Vl’s such as “Move Read” and “Surface vi”. “Move Read” contains the program for the 140 stepper motor drivers and for reading the reflected power. Surface vi contains all of the components of the simplex method discussed in the previous section. This is a major program which calls several VI’s such as indicated in the program hierarchy. Level 3 is the mode selection controller which involves several levels of nested loops or sequence structures. This program implementation follows the logic in Figure 5- 16. It is important to note that only pertinent details of the LabView program have been presented. The purpose is not to serve as a tutorial but to bring forth the significant highlights of the work. To use this program does not necessarily require knowledge about LabView since most of the interaction would occur at the front pane which is self explanatory. However, to modify or to understand the program code does require basic understanding of LabView. 5.7 Diagnostics System Automation 5.7. 1.1 Hardware Automation The sweep oscillator is a low power multi frequency radio frequency power source which operates in an output power range of 10-20 mW at a variable frequency range of 1.7-4.3 GHz. It is used for the low power swept frequency analysis of the empty cavity to locate theoretical modes. In this analysis, the reflected power from the cavity is rectified by a crystal detector and displayed on the Y-axis of an oscilloscope trace the quality(Q-curve) curve of a mode. Figure 5- 16 shows the device diagram for the low power diagnostics circuit. The difference between this circuit and the processing circuit is that the microwave power supply is a variable frequency source, sweep oscillator and not the fixed frequency source, Magnetron. Also, since the power levels are at milliWatt levels, the reflected power from the cavity is not attenuated but directly read by the power meter and then sent to the computer via the AID board, while the swept 141 fl Ewen e< IJ mu: 2.5. xom REEL»... + > m .555... < .855... 2.9... w......:eU \ .— 5.2.5.28...— uoaoEOE—Seom 8.... agree .250... 2......— w......=eU .53.... d .52.... 4 .32.... Q .39.. 5.5.3 8n... omen ._ .53.... d 5...... _ .5933 8a... u..........m f \ 5.2.0 LOX—=25 cousem .52 eoznm 3352.82 503mm 858.555 iagnostic System Figure 5- 168 Low Power D 142 frequency is controlled via the GPIB interface. See Appendix A for detailed discussion on GPIB interface. In manual Operations the swept frequency was controlled from the sweep oscillator and the reflected power was attenuated and then sent to an oscilloscope. Thus, the oscillosc0pe was used to track both the swept frequency and the reflected power. 5.7.1.2 Diagnostics system software development The objective of the diagnostics system is to characterize the microwave processing system, which includes the cavity and the external circuit for theoretical cavity modes. As was previously discussed in the hardware chapter, a low power multi-frequency source is used to study the resonance of the system at the excitation frequency. In the manual Operation the resonance data is displayed on an oscilloscope as a plot of frequency versus reflected power, where frequency is displayed on the x-axis. In the automation of this procedure, data acquisition at resonance is automated while the decision for cavity adjustment for tuning is not automated. This is due to the low power values (0-10 mW) that are output by the sweep oscillator and the limited resolution of the sensing equipment and data acquisition system. To have a fully automated system the low power circuit has to be augmented by the addition of sensors with resolution for low power values or amplification of the signals before being sensed and sent to the data acquisition board. The software for the low power data acquisition system uses two different modes of instructions, one for a GPIB interface to control the swept frequency and analog-to-digital instructions to acquire the reflected power data. Input to the program are the sweep oscillator address, swept frequency rate, minimum frequency to scan and maximum frequency to scan. 143 Given this input the program calculates the number of times to iterate the data acquisition loop or the number of data points to acquire. A string of frequency is then sent to the GPIB write driver at the instrument address of 19 to set the swept frequency. Once this instruction is completed the data acquisition board is queried at the reflected power channel to read the reflected power value at the set frequency. This instruction is repeated until the number of iterations set by the program is completed and then the data is displayed on the front panel of the VI. The actual output is shown on the tap and an actual oscilloscope photgraphic image is shown on the bottom for comparison. The data can also be available for further analysis in Excel. The significance for the automastic generation of the Q-curve is that it provides the ability for automatic on-line cavity diagnosis and dielectric analysis. In further development, cavity frequency shift calculation and Q-curve calculations were developed for a unimodal curve. Further developemnts must be made in the developemnt of methods for handling multirnodal curves. It is important to note that the subroutines for interfacing with the GPIB where provided with the GPIB board. Hence, the extent of the software development was the structuring of the device drivers and other available subroutines. 5.8 LabView Software Interface with Knowledge-Based-System Planner As previously mentioned, another software program was developed for the purpose of interfacing with a KBS-planner. Details of this work are discussed in Appendix C but the software development issues are presented. The program flow logic for interfacing with the KBS-planner is shown in Figure 5-19. Note that it uses the sub-programs for power controller, mode tuner, and data acquisition that were developed for the processing system discussed in the previous sections. What is 144 different however this program is that the decision making now resides in the planner. The acquired temperature, input power, reflected power, cavity length and probe depth are input to the KBS planner. The output from the KBS-planner are the power level and modes. Hence, the LabView program functioned as a low level controller which takes the control setpoints from the planner and maintains them. It is also important to note that the interface between the planner which was implemented in SmallTalk, and the LabView program was done using LabView subroutines available in the file management library. Hence, this was another demonstration of the versatility of the LabView software program and its capability to interface with different software platforms. Details of the programs and philosophy of this work is presented in detail in Appendix C. 145 [ KBS Planner J Power Level Mode | Power Controller I i [ Mode Tuner I I Figure 5- 19 Logic for Interface with planner (Figure 1-2) CHAPTER 6 APPLICATION TO CURING 6.1 Introduction In the previous two chapters the empty and loaded cavities were characterized in order to understand the excitable modes using theoretical and empirical approaches. In this chapter the empty and loaded cavity characterization results will be applied to the automatic curing of graphite epoxy composite materials. The goal of these experiments is the verification of the automated system for controlling the curing process by selecting modes, tuning, and controlling the sample temperature and temperature gradient. Using the characterization results, the heating modes are grouped into edge and center heating modes. Thus, the uniform heating criteria is set to be the minimization of the temperature gradient between the center and the edge temperatures first, and then between the edges. The graphite epoxy system was used because of the complex nature by which it interacts with the electric fields. The graphite fiber component of the composite lends a highly conductive characteristic to the composite, making it highly lossy and very responsive tomicrowave heating. Thus, it provides an interesting medium for an extreme case by which the automated system can be verified. 6.2 Curing Experiments In the curing experiments a graphite epoxy composite prepreg, Hercules AS4/3501-6, was layed-up into 24-ply and 48-ply, unidirectional, 2 inch square parts. The sample was cured at setpoint temperature of 160 ° C for a total processing time of 90 minutes. The lay-up method follows that used in the autoclave processing but, unlike the 146 147 autoclave layed-up samples the part was not placed in a vacuum bag since pressure was not applied in the microwave process. High temperature (200 °C) lay-up material was used as compared with the polyester based materials which burn at temperatures less than 160 0C. A Teflon mold which has been specifically designed for the microwave cavity was used to contain the sample during curing. The Teflon mold was machined out of a solid block to form a male and a female unit, which were fitted with pegs to lock the two pieces in place. Sample surface temperature was measured using fiber optic therrnornetry, with the temperature probes inserted through the top of the Teflon mold and secured in place with autoclave tape. Sample placement was such that the coupling probe was perpendicular to the fiber direction and the sample was elevated to a height of 1/8 7L. Five temperature probes were used which were placed at equal distance at the four corners of the sample and one in the center. The modes or cavity lengths used and the estimated preferential heating sites are listed in Table 6—1. The probe depth was always set to be less than 25 mm in order to minimize the near field interactions with the sample. These modes represent the tunable modes for the sample loaded cavity(see chapter 3). The preferential heating sites were estimated from the loaded cavity characterization results to be those shown in Table 6—1. ICavity Length (cm) I 8.9 I 12.7 I 13.5 I 193 I 22.2 I 23.3 I [Heating Preference I Center I Edge I Edge I Center I Edge I Edge I Table 6- 1 Loaded Cavity Resonant Cavity Length and Estimated Preferential Heating Sites 148 6.3 Results and Discussions 6.3.1 24-ply Sample 6.3.1.1 Mode Selection Figures 6—1 through 6-6 show the curing results for the 24—ply and 48-ply samples. Results are shown in the form of temperature profile, power profile and cavity length and probe depth as function of time. For the 24-ply sample the overall processing results are listed in Table 6-2. The center temperature is the dashed line and the rest are the edge temperatures. The temperature profile in Figure 6-1 shows several dips which are indication of mode switching periods. These dips show the sensitivity of sample heating to the electromagnetic coupling and mode switching times. Figure 6-2 shows that during the first 10 minutes, the mode selection program used all S-modes , but stayed in the center heating modes the longest This is an indication that the sample center was the coolest during the first 10 minutes of the run. Between 20 minutes and 30-minutes only the center heating mode was used at cavity length 19.7 cm. This was a good choice by the mode selection program which is indicated by a uniform increase of the sample temperature profile from 60 °C to 80°C. Between 40 minutes and 50 minutes, all the modes were used and none was favored. This was the region where the sample temperature was in the reaction region. It seems as though the switching times increased as the reaction temperature was reached. This is an indication of the program keeping up with the reaction front such that the sample is uniformly cured. After 77 minutes or 30 minutes in the reaction zone, the dips in the temperature profile seem to be minimized although the program is selecting and using different modes. This can be interpreted as advancement of the curing reaction. 149 The reaction was controlled for 90-minutes, since this was a predetermined time when complete curing can be successfully achieved(Wei, 1994). 6.3.1.2 Power Control Power control did not start until 50 minutes into the run. During initial stages of the run the power was kept constant at 100% output. The dips in the output power is a characteristic of the power source and not the controller. Thus, the maximum output power was an average of 80 W. The power controller regulated the power such that the sample temperature never overshot the setpoint of 160°C. Note that after 70 minutes the response of the controller became very rapid and the power level was reduced and the sample temperature remained to be in the processing window. This could be an indication power control to compensate for reaction exothenn 6.3. 1.3 Tuning The reflected power was considerable higher at approximately 20% of the input power, which was due to the restricted probe depth requirement in order to minimize the near field effects from the probe. For completely minimized reflected power a probe depth of 30 mm or greater is typically required for this system. However, with the long probe depth the near field effects from the probe are increased and heating patterns become unpredictable. Thus, the limited probe depth requirement restricted the optimum tuning capability of the system to the best that it can do at a maximum probe depth of 25 mm. The tuning and mode switching times were on the order of a minute or less with the cavity adjustment time being the controlling factor and the tuning time on an average of 20 seconds or less. Typically, the tuning algorithm can a achieve a minimum reflected power of 10% or less of the input power. Theoretically, a minimum reflected power requirement for 150 tuning is zero or very close to zero. One of the consequences of partial tuning is un- focused energy which could have also contributed to the dips in the temperature profile. With the limitation on coupling probe depth, a more predictable heating profile was achieved across the sample although the heating rate was slow. This condition can be remedied by using a higher input power to achieve a higher heating rate without sacrificing heating uniformity. Figure 6-2 shows that the coupling probe depth remained constant during major periods of the cure cycle, although several different modes were used. This is to note that coupling probe adjustment were not significant in tuning the cavity during the cure cycle. The only periods that the coupling probe adjustments were required was during the pre- reaction period, between 30 minutes and 40 minutes which corresponds to a sample temperature of 80 and 120°C. tem. 1n ttIme < SEC unrn time < sec ycunng 6.3.2 48-ply Sample ' 6.3.2.1 Mode Selection Figures 6-4 through 6-6 show the curing results for the 48-ply graphite epoxy composite material. The same modes and processing conditions set for the 24-ply material were used and the results are summarized in Table 6-3. Figure 6-4 shows the curing temperature profile for the 48-p1y sample. The highest temperature profile is the center temperature and the rest are the edge temperatures. Note the absence of dips that were present in the 24-ply sample. 151 During the heat-up period, the estimated two center heating modes were selected the most since the center temperature had the lowest value. This was similar for the 24- ply sample. The mode selection controller continued to use these 2-modes since the overall temperature was maintained within 20 °C, which was the criteria for uniform heating for this sample. By 14 minutes into the run the cure temperature of 160°C was reached. Between 14 minutes and 30 minutes, the edge heating modes were selected in order to push the edge temperatures into the processing window. During the whole curing period the edge heating mode at a cavity length of 23 cm was selected between 30 minutes and 80 minutes except for at 50 minute into the run. This is a clear indication that heat loss at the edge of the 48-ply sample was significant. 6.3.2.2 Tuning Figure 6—4 shows the input and reflected power profile for the 48-ply material. During the heat-up period the tuning was very efficient, and the reflected power of 1% of the input power or less was achieved even with the limited coupling probe depth requirement. As the cure period proceeded tuning became less efficient with a reflected power of approximately 20% of the input power. This could be due to the coupling nature of the more massive 48-ply material. During the period of 30 minutes and 81 minutes, the reflected power was approximately 20W, and between 80 minutes and 90 minutes it was 0 W. 6.3.2.3 Power Control Power control was more interesting for the 48-ply sample than the 24-ply sample. Because the heating response of the sample was very rapid, the response of the power controller was very rapid. This is a very interesting result which shows how the power controller complements the mode selection controller. Note that every time that a new mode is selected a new region is preferentially heated while the hottest region is cooled. 152 Hence as the measured input to the power controller is constantly changing as new modes are selected. However, in the region between 50 minutes and 80 minutes, only one mode was used. Hence, the input to the power controller was fixed which caused the rapid and instantaneous response of the controller. Similar response was also observed for the 24- ply sample between 75 minutes and 90 minutes where a single mode was used. It is also important to note that, between 81 minutes and 90 minutes, although the power was reduced to an average of 40 W, the sample temperature was maintained within the processing window. This is a good illustration of the compensation for reaction exotherrn in the microwave system. rate mrn. rme IO I‘BZlC SC I minutes tern . I < vera < in l trme unrn time < sec p y curing ts 6.4 Summary and Conclusions The automated system was successfully demonstrated in the curing of 2-inch . square, 24-ply and 48-ply, unidirectional graphite epoxy composite materials. Temperature gradients of 10 ° C or less and 20 ° C or less were maintained for the 24-ply and the 48-ply samples, respectively. Mode switching and tuning times were on the order of a minute, which was constrained by the physical cavity adjustment times. The 48-ply sample was more forgiving and was not significantly affected by the cavity adjustment delay times. The actual output power was less than 100 Watts and the reflected power from tuning was high, approximately 20% of the input power, which was due to the restriction 153 on probe depth and the long probe depth requirement for Optimum tuning for these materials. The estimated heating patterns and mode locations in the cavity were sufficient in the selection Of modes to achieve edge-center heating uniformity and in the location of modes in the cavity. Input power was successfully controlled to maintain the sample temperature at the curing setpoint of 160 ° C. In general, as energy was coupled into the sample, the sample heated well, and as energy was not coupled into the sample temperature decreased rapidly. The center heating mode was used most frequently than the edge heating modes. This could suggest that heat loss from the central part Of the sample was more significant for the 48-ply graphite / epoxy sample. The heating characteristics Of the 48-ply material is very different from the 24-ply sample. The time delay in mode switching did nOt have a significant effect on the heating Of the sample. There seemed to be enough thermal mass established in the material such that it was more forgiving to cavity adjustment delay times and edge heat losses. Edge heat loss was more significant and thus edge heating modes were preferentially used. Finally, using these two samples which characteristically heats differently, the developed program was used to successfully cure them. These results indicate the versatility Of the automated system and its capability to adjust for different processing requirements. 154 58%. :3... 329.5 :93 .530 .8 3:2.— o.=.a.o..Eo.—. To 0...»... Essafi 2.3 ndm 3...... new 3.9. 3.... S... h w 0 n .1 .r . a -- mammuflnw - .. ‘ 2 U sank. ....... r 1.1! ~.. . .. 2:. i..- .x . Elli .. n 3:. 9. Fill: _ nTIII .. 8 - 53.... .. - .. will... 3... .. cm x x, .. 8. .1“ . on. _r , c3 c8 cw— (3) madam. ram :3... 22%... be... 9.2.6 a. 5.2. 22.. a... 5.5.. as... 2. 2...... 155 35.8.... 2.3 ndw 8.2. 03 9...: 3.: 5.: w i. x .r . x c z . n n . .uu . u... "I u . _. . .I _. _ : .n H. _. n c —- .c _ n“ - _- .. .n u . - _. . . . z. i .u. u _. can". ._ .- .. ,- . ~ . .. . . an. . .- __.&..n........w.lll._l;é-_ . . 7..-. . . . cm N — — (mm) mdao aqua Am) 10307! Kama N 156 .38.... \ 0.29.5 3.5.8262... 5.93 9.2.5 .8 .03.... 8.00:0... .2... =5... m-.. 0...»... .. 3.5.0.5.... :dw “.mw was... 38.4.11 ..... . .. Alli. B .J u. m Wmtlcoi‘o‘dfid. ‘30:... *uath..0\o.’foqoao$p‘ M m o, .QoouucoI I 0“! fl Cum 5... .0 Ccoooon---“ Q00. .-------------. .----,_---------4 :cm .. an .. cm -8. (m) mad muse rm and"! 157 2.8....50 $25 20...... 0.2920 2..-: »...:..U 8. 050... 0.30.0080... v-0 0...»... 3.8.0.5... no.8 hog-w 504% he. . w w.- mmdh w. _ n 5. .NM ac... _ . . . .1 . x . . . . c 8."... ....... . 02".. ....... . cu n... I .. 9. a. v n Ell _ Elli! .. 8 i---.._-..M....-....-ul . . .. x. .. 8. 3...“... .- GN— ) o. ................. ----...: firm-F“... ....-.n...%-\ :::-:3. .JJr-rr-t. .u r nnnnnnnn no nnnnnnnnn non-o-.. .. orurYrrr. ..... urn-18— aha-F a». (3) murders; 158 .08.... 20...”. 0.39.... 2..-: ».......U .0. £00m. 0...... 0.... ._.»..0.. 5.2.0 n-.. 0...»... we...» no...» no...» 3.. 2.8.05... » an an... n. _ n 2.3 as... b n b < - d 1 fi . - h 4 1 JG - .. .52... 2:-.. el . [1 V5 0 (mar) who «101d pm: (mo) mfiuaq tip-23 n —n O N nu 159 0:08:80 58%. :0...”— 0.E..«.O 2..-? «5.20 .0. .03.... 8.00:0: 2.0 .03.... 3...: ob 0.53"— 5.3 5.; A5525? .2. ands. n._n 2 .mm -c---’ W 7.... p 4 . m... . _. .fi»? . LU :— '3 cm : an : ch 1. c” 8— (M) Jamod CHAPTER 7 SUMMARY AND CONCLUSIONS 7. 1 Introduction The purpose of this work was to advance the fixed frequency single-mode resonant technology for the processing of composites by bridging the gap between device and process. This was important in order to realize the technical potential of the microwave processing technology as a viable alternative to thermal processing. In the single-mode resonant cavity, the electromagnetic modes are discrete and their interactions with composites are complex, non-linear and time variant. In general the development of complete mathematical models to accurately describe the dynamics of the electric field and heating behavior inside the cavity is complex and computationally intensive. This is a major source of difficulty which is manifested in the lack of advancement in the state of technology of the single-mode resonant cavity. In the previous state of technology, the single-mode resonant cavity was operated as manual device that is impracticable as a process. The hardware consisted of four basic components: 1) tunable cavity with gear drives for manually adjusting the cavity length and probe depth, and micometers for measuring the cavity length and probe depth; 2) an . external circuit with major components which included a microwave power source, power meters, and a dummy load to absorb reflected power; 3) fiber optic thermometry for invasive temperature measurements; 4) computer interface for automatic data acquisition of input power, reflected power and temperature for post-analysis, and the onIoff control of a switch to direct the microwave power away or to the cavity. There were two different independent external circuits for processing and for diagnostics purposes. 160 161 In processing, the microwave cavity was operated as an open-loop system where a seasoned operator was the necessary link to close the control-loop. A typical processing activity included; 1) the continuous tuning of the cavity by the manual rotation of dial knobs to adjust the cavity length and probe depth to minimize the reflected power; 2) the frequent selection of new cavity modes by carefully adjusting the cavity length and probe depth to new locations, while tuning the cavity and manually modulating the input power to achieve uniform heating; 3)temperature control by the automatic onIoff control of an electronic switch to direct the microwave to or away from the cavity. As such, the cavity was operated as an independent device from the external circuit, and processing results varied from one operator to the other and optimization was difficult at best. In this work, the single—mode resonant cavity was automated in order to advance it as a viable process. In the automation, a control system was designed and built in addition to the development and implementation of a set of sophisticated and comprehensive control software programs for controlling the curing process in the cavity. These control programs combine traditional and non-traditional control methodologies. The control software programs were developed for mode tuning, mode selection and power control which were constructed independently and then integrated to form the overall closed-loop feedback control system. The diagnostics system was also automated to provide for automatic empty cavity characterizations and for automatic dielectric analysis of materials inside the . cavity. In the sections that follow, the control system for processing is summarized under the following headings; control software, automation hardware components and application results. The automation of the diagnostic system is summarized in a dedicated section, followed by a summary of discoveries and developments from the course of this work that would lead to further understanding and utilization of this technology. 162 7. 2 Automatic Control Software A closed-loop feedback control system was deve10ped which included software and hardware for physically controlling the microwave cavity and external circuit as an integrated process for curing composites. Mode tuning, mode selection and uniform heating and power control were the major components of the control software. Other supporting programs were developed for the stepper motor drivers and for data acquisition. These suporting programs are presented in apprecaible detail in appendix B. In processing, electromagnetic modes were selected using the mode selection controller, and the modes were tuned using the mode tuning controller, while controlling the input power to achieve uniform and controlled heating. The control software programs were structured in series in the following order; data acquisition, power controller, mode selection and uniform heating controller and mode tuning controller. This means that only one controller was active at any one time, and each controller has to complete the control task before the successive one could be activated. This also means that the completion of one controller activates the other controllers. Other elements of the control software were; data acquisition and storage; real-time tracking of process variables in the form of strip charts, color maps and spreadsheets; safety locks such as temperature alarms, power alarms, and multiple shutdown levels. . 7.2.1 Mode Tuning Software Program Using a non-traditional control approach, a mathematical 2-dimensional simplex method was used to construct the tuning control software. In this method both coupling probe depth and cavity length were adjusted simultaneously to tune the cavity. Application of the simplex method to mode tuning and the simultaneous adjustment of the cavity length and probe depth to tune the cavity are both novel in this approach. This method only required information about the cavity length, probe depth and reflected power and not the electromagnetics inside the cavity. As such, it was simple to program and yet more 163 efficient in tuning the cavity than the manual and univariate methods. Using this method tuning time was typically on the order of a nrinute or less which was comparable to manual tuning, but mode tuning was highly reproducible which is an improvement from the previous state of technology. One shortcoming of the simplex method is the dependence of the rate of convergence on the proper scaling of the cavity length and probe depth adjustments beforehand. These scaling factors have to be determined experimentally following similar principles as applied in controller tuning. Thus, in order to optimize the tuning rate of convergence the scaling factors must be optimized. 7.2.2 Mode Selection and Uniform Heating Control Software Program Mode selection and uniform heating control software was required to physically select the appropriate cavity length and probe depth corresponding to a desirable heating mode. The uniform heating controller was developed from a non-traditional control methodology using empirical correlations to construct the heating functions and electric field characteristics of the loaded cavity. Chapter 4 is dedicated to the development of these correlations. Uniform heating was characterized as the maintenance of the spatial edge and center sample temperatures within a 5 °C temperature band. The loaded cavity modes were grouped into center and edge heating modes from the loaded cavity characterization results. Uniform heating was achieved by the selective switching between these electromagnetic ' modes to preferentially heat the center and edge regions of the sample, to distribute the electromagnetic energy across the sample in an optimum manner. Ideally, the center region of the sample was initially heated preferentially, until a gradient of >5 °C was obtained between the center and edge temperatures and then an edge heating mode was selected and the procedure was repeated. The controller was found to be sluggish at times where several mode selection iterations occurred before the correct one was selected. This was in part due to near-field 164 and time-variant complex electromagnetic interactions which could have caused variability in the electric field patterns of the predicted modes. Thus, indicating that the heating functions and the mode selection criteria can be Optimized, and the loaded cavity modes may not be neatly grouped into center and edge heating modes as was done in this controller. Another factor that could have contributed to the sluggishness of mode selection was the mechanical cavity adjustments involved in the mode selection. Although the adjustment time was 10 inches / minute, which was a 67 % increase in speed from the manual system it may not have been rapid enough. The uniform heating method used in this controller was found to be sufficient in the demonstration of the concept, however a better approach would be to develop more accurate heating functions using either mathematical models or the rigorous empirical methods proposed by Fellows(Fellows 1994) in order to optimize the mode selection criteria. One notable point is that uniform heating in the single-mode resonant cavity is not static. It is complex and does not only depend upon the ability to rapidly select and tune to a mode, but highly depends upon the understanding of the time varying characteristics of the heating field patterns. 7. 2.3 Power Control Software Program Power control software program was developed to physically modulate the input ' power to maintain sample temperature within setpoint limits. Power control software was developed using traditional proportional-integral-derivative (PID) methodology, and the microwave power source was modified to allow for analog adjustments of the input power. Since several temperature sites were measured, the input to the controller was the highest temperature value. Using this controller, thermal overshoot from the reaction exothermic was effectively controlled and temperature was maintained within 2°C of the setpoint. 165 This demonstrated the controllability of temperature excursions in the single-mode resonant cavity which was difficult to do in the previous state of technology. In the previous state of technology, the concept of temperature control was the onIoff control of an electrical switch to direct the microwave power to or away from the cavity. While the power was directed away from the cavity it was directed to a dummy load as wasted power. Hence, in addition to the demonstrated technical advancement of the developed controller, it was also an economic advancement in the use of the microwave power. 7. 2.4 Data Acquisition Interface and Control Software Platform An element of the automation was the hardware control system development which included the microwave cavity, microwave circuit and the computing and data acquisition platform. A Macintosh computer running a LabView software development program was used for computing and software development, respectively. The choice of the computing platform was motivated by the capability to interface with a knowledge-based system unit, capability to perform simulations and the ease of use. However, due to the recent technological advancements in computing software the work developed will be portable to other platforms with similar ease in operation. The LabView software platform was found to be user friendly both in program development and in application. Additionally, because ' LabView contains comprehensive data acquisition drivers and analysis subroutines it was found to be flexible and proficient and easy to maintain. 7. 2.5 Data Acquisiton Software program Although the LabView software platform contains data acquisition drivers, a program had to be developed for acquiring the process data The program essentially included sampling rates, scalin g factors for the different instruments and data manipulations 166 functions for minimizing noise in the acquired data. Data acquisition components are presented in appreciable detail in appendix B. 7. 3 Automation Hardware 7. 3.1 Cavity and Circuit The manually operated microwave circuit design was essentially modified to accommodate automatic data acquisition, and the microwave power source or magnetron was modified for automatic analog manipulation. The core of the hardware automation was the design and fabrication of the cavity with mechanized drives. The overall cavity design was not novel and had been previously developed with either an axial or a radial mounted coupling probe(Asmussen 1987). What was novel about this cavity design was the mechanized drives and the implementation of dual mounted coupling probes at the axial and radial positions. The mechanized drives design was based upon a belt and pulley system in which a stepper-motor was used to drive a ball screw which resulted in smooth adjustments. Additionally, the use of precision graphite bearings rather than spur gears eliminated the binding problems that existed in the manually operated cavity. In the operation of these drives, it was shown that one shortcoming in the design is the sensitive to alignment where proper alignment was necessary for optimum performance of the drives. This is a problem ' which can be easily remedied by the design of precision brackets to keep the drives aligned. Although the mechanized cavity diameter was the same as the manual cavity, the length of the cavity was 42 cm which is 1.5 times the length of the manually operated cavity. This increase in cavity height allowed for the excitation of 12 additional modes which increased the processing flexibility for multiple mode applications. Another element of the cavity design that was novel, was the addition of linear motion potentiometers for electronic and local cavity length and probe depth measurement. 167 7.4 Verification of Control System To verify the integrated automated system, a 2-inch square, 24—ply and 48-ply, unidirectional graphite epoxy composite materials were cured. Chapter 6 is dedicated to the discussion of these results. Sample specific inputs to the control program were; the matrix type from which the cure temperature setpoint was automatically selected; cure duration time; and heating modes grouped into center and edge heating modes. The automatic curing results showed that benefits due to automation versus manual operation are; 1) 2) 3) 4) 5) A 63% decrease in mode switching times due to the precision mechanized drives and the effective mode tuning methodology. This is notable in that sample heating interruption times and thus sample cooling times during mode-switching were minimized which resulted in the improvement of sample heating uniformity. A 50% decrease in sample temperature gradients within a 5 ° C bandwidth, and thus increased sample heating uniformity. This was clearly due to faster mode switching times, efficient power control and the mode selection criteria for mode switching. Proficient and repeatable mode tuning and elimination of phantom modes which minimized processing variability. Efficient control of exothermic reaction temperature excursions where temperature was maintained within +/- 2 °C of the setpoint. Efficient use of microwave power where only reflected power was directed to a dummy load which was always less than 20% of the input power. Other benefits that are noteworthy; 1) User friendliness of the process control software, 168 2) Real-time tracking of the process trends using strip-charts and other visual tools on the computer, 3) Integration of data acquisition and control on a single platform, 4) Accurate documentation of the processing results for further developments to eliminate the invasive sensing methods, 5) Availability and easily accessed processing data in a spread sheet form, and most importantly, 6) Convenient turnkey processing characteristics as in the operation of thermal ovens. 7. 5 Diagnostic System Automation As previously mentioned, the diagnostic system is an independent external circuit which consist of similar circuit components as the processing circuit, except for that the microwave power source is replaced by a sweep oscillator. Additionally, there is an oscilloscope for measuring the input power and reflected power in the form a power absorption curve. The automation of the diagnostic system was significantly different from the processing system in that a GPIB (general purpose interface board) interface in addition to the analog-to—digital interface used in the automation of the processing external circuit were used. This was necessary since the major component of the low power diagnostic ' system was the sweep oscillator which has only GPIB interface capability. A mechanical switch was added between the processing and the diagnostic external circuits which would potentially allow for automatic on-line dielectric analysis during processing by switching between the two independent external circuits. Essentially, automation of the diagnostic system replaced the need for the oscilloscope and the manual calculation of the cavity quality factor, which enabled the automation of dielectric analysis in the single-mode resonant cavity. One shortcoming of 169 the program was that it was only capable of calculating cavity quality factor for unimodal power absorption curves. Hence, further development is required to complete this work by developing a more robust algorithm for calculating the cavity quality factor for non- unimodal power absorption curves. 7. 6 Understanding to Enhance Utilization of Technology In the course of this work several high level global understanding of the microwave processing of materials was developed or discovered which enhanced the utilization of the technology. These areas of understanding can be realized in hardware design, processing methodology. and the application of knowledge-based system technique to automation. 7.6.1 Dual coupling In the design of the mechanized cavity, as an added feature which was not required for automation, the mechanized cavity was equipped with novel axial and radial mounted coupling probes. As was previously mentioned typically, either the axial or radial mounted probe configuration is used and not both. Empty cavity characterization results showed that similar modes can be generated for either probe mount configurations except for at the cavity length corresponding to a degenerate mode. An interesting finding was that, only at the cavity length corresponding to the degenerate mode can simultaneous coupling from both axial and radial mounted coupling probes be used to achieve heating. At the other cavity lengths as one coupling probe coupled the microwave energy into the cavity the other coupled it out of the cavity and the sample was not heated. Theoretically the side mounted probe should preferentially excite TE-type modes while the axial mounted probe excites TM-type modes. This may suggest that at the degenerate mode, each probe preferentially excites standing waves that are orthogonal to one other and do not destructively interfere with one other. However, at the 170 non-degenerate modes, both the axial and radial mounted probes excite similar modes that may be out of phase with one other and cause destructive interference; which was indicated by one coupling probe coupling the energy out of the cavity. This would be a hardware related problem that can be remedied by phase locking the microwave sources attached to the axial and radial mounted coupling probes. The heating pattern that was observed using the degenerate mode, was uniformly distributed across the sample surface and resemble the superposition of the degenerate modes. What is notable about the simultaneous coupling method is that it provides for an additional methodology for achieving uniform heating in adjuvant to the mode-switching method using only the radial mounted coupling probe. Results of this work are discussed in appendix. The axial mounted probe is also significant in that it can be an alternative to the radial mounted probe in order to address the near-field interactions experienced from the radial mounted probe. Cavity characterization results from chapter 4 indicate that the near- filed interactions can alter field symmetry which can result in unpredictable field characteristics. 7.6.2 Coupling Probe Effects Even though the coupling probe is typically designed to be 40 mm long, it was found that at greater than 20.0 mm it can interact with the elecuic fields in such a manner that field behavior become highly unpredictable. This was said to be due to near-field ' interactions that were more significant with longer coupling probe depths. Hence during processing, the coupling probe depth was limited to less than 20.0 mm in order to minimize the near-field effects. The coupling probe alignment was also found to have an effect on the field symmetry. The more the coupling probe was not alligned, the more non-symmetrical the field pattern. This was found to be a hardware problem that can be resolved by properly aligning the probe mount with the cavity axis. 17 1 Sample placement height with reference to the coupling probe height was found to affect the field pattern inside the cavity. As the sample height was close to the coupling probe height unpredictable field behavior as was observed for the long coupling probe depths was also observed. Sample placement was optimized to be 1/8 of the freespace wavelength away from the coupling probe. Another interesting finding was that during processing, once the cavity was tuned to resonance for a unique mode, cavity length adjustments alone were enough to tune the cavity to resonance even when new modes were selected. This was observed during both graphite / epoxy and vinyl ester / glass composites processing. 7. 6.3 Loaded Resonant Cavity length Typically, as a large and highly lossy material is loaded into the cavity, the electric fields become so perturbed that the calculated empty cavity solutions become no longer valid. As such loaded cavity modes become difficult to characterize. The empty cavity characterization results from chapter 4 showed that, the change in the loaded resonant cavity length from the empty cavity, for a typical 24—ply composite with a dielectric constant ranging from 2 to 4 was approximately 3.0 cm. This approximation allowed for the characterization of loaded cavity modes which was used in the mode selection controller to achieve uniform heating. It was found to be a sufficient estimation for lower order ' modes and invalid for higher order modes. This may be due to the more unique ordering of lower order modes and the more random ordering of higher order modes. 7. 6.4 Effect of Tooling In the loaded and empty cavity characterization for the development of the uniform heating controller, information about sample placement, near-fields, coupling probe alignment and depth, resonant cavity length and Teflon material were discovered. As the 172 size of the Teflon block was increased the fields were found to be pushed more and more to the boundaries of the block What is enlightening about this finding is that even though the Teflon material may be nearly electrically transparent, it still can have an effect on the electric field inside the cavity due to the volume. This clearly suggests that tooling effects cannot be ignored and must be actively studied from the point of design optimization, to other alternatives which would integrate the microwave cavity and the tooling piece as one unit, i.e. the part shaped cavity. 7.6.5 24-p1y versus 48-ply curing In the automatic curing of the 24-ply and 48-ply graphite epoxy composites, it was observed that the 48-ply sample was more forgiving to long mode switching times than the 24-p1y sample. This was indicated by sharp dips in the temperature profile for the 24-p1y sample during mode-switching times which was not present for the 48-ply sample. These results are discussed and depicted in chapter 6. The important finding from these results is the implication that the mechanical tuning technology may be more applicable to the processing of thick-section samples than to thin section ones in which the rate of heat loss is significant. Hence, frequency switching which could potentially provide for faster mode switching times would be more applicable to the processing of the thin-section composites. 7.6.6 Theoretical Field Pattern Plots In the application of the empty cavity solutions to characterize the empty and loaded cavity, a method was developed for generating theoretical electric field patterns in the form of density plots for the empty cavity. This method uses easy density plot functions in a software program called Mathematica. Details of these plots are discussed in appendix A. The generation of the electric field patterns in the form of density plots is unique in that contour plots are typically used which are difficult to understand and the not all the field 173 pattern plots are available. These plots were used in the characterization of the loaded and empty cavity modes and were found to provide more comprehensive information than the contour plots. 7. 6.7 Application of Knowledge-based System to Automation The lack of fundamental process models and the complexity of the domain in this technology, motivated the application of knowledge-based systems technique using a generic task approach for the overall automation. Results from implementation of the knowledge-based system technique for low level control showed that, it was more effective for high level tasks such as planning than for fundamental control tasks. This result motivated the development of the global microwave curing control task into two parts; a low level traditional unit and a high level knowledge-based system unit. Hence, a knowledge based system technique was applied in a collaborative research effort. The knowledge-based system planner was developed as a core dissertation work by Decker(Decker 1995) which was interfaced with an automated system developed as a part of this dissertation, specifically for interfacing with the KBS system Output from the planner to the automated system were electromagnetic modes and power levels required to control the curing process to achieve uniform heating. Input to the planner from the automated system were the sensed process variables. Hence, the ' automated system had capabilities for acquiring data and carrying out control tasks that were requested by the planner, where the planner was the control decision making unit. The noteworthy element of this work is the interfacing capability between the knowledge-based system planner that was implemented in a software language called SmallTalk, and the automated system that was implemented in LabView. The successful demonstration of this system in curing experiments is a significant accomplishment in the application of a knowledge based-system planner in the real-time control of a complex 174 process, and motivates new directions in process automation. It provides the potential for capturing and representing experiential knowledge which typically resides with operators. This can be further advanced to knowledge retention and learning with the intended goals to eliminate the invasive sensing methods in the microwave processing technology. Chapter 7 is dedicated to the discussion of this work. 7. 7 Global Conclusion In conclusion, the fixed frequency, single—mode resonant microwave concept was fully automated and demonstrated as an automated “1“ generation prototype process”. The benefits due to automation clearly indicate a significant step in the advancement of the single-mode resonant technology as a viable automated process. In the course of this work, high level understanding to enhance the utilization of the technology and potential applications were developed. Despite the extensive amount of research that has been ' expended in the development of the “1“ generation prototype process”, there are still several issues that must be addressed to completely advance this technology to commercialization. Some of these issues include economics feasibility, the development of a user friendly microwave processing unit, and most importantly specialized tooling, elimination of evasive sensing methods and scale-up. Whatever the commercialization goals, whether it is curing, drying, heating, ' dielectric analysis, plasma applications, batch or continuous processing, practical and convenient process characteristics are paramount As such, the completed automation work is a significant enabling step in providing the foundational elements to realize the viable potential of the single-mode resonant technology, where further research can be pursed. Finally, although this research program was comprehensive in scope, its goals were significant and successfully directed toward the technical advancement of a viable 175 process with numerous potential applications, and the motivation of new research challenges. CHAPTER 8 RECOMMENDATIONS AND FUTURE WORK 8. 1 Automated System The automated system developed in this work is the “1" generation prototype” which was shown to significantly advance the single-mode resonant system as a fully automated process with a viable potential for commercialization. Although significant amount of research has been expended to evolve the technology to this stage, there is still work to be done to optimize the technology in order to further advance to commercialization. In general, the automatic control system can be modified to achieve more optimum control. This may be done by expressing the control elements in the frequency domain rather than in the physical domain as was done. This would make the control system more robust and easier to maintain over the years. However, achieving this goal is contingent upon the generation of accurate and complete process functions which can be done either from exact solutions or from “good” estimations. With the contributions from this work and other developments such as the results from Fellows(Fellows, 1995) this can eventually be done. The uniform heating controller can be improved by either optimizing the heating ' knowledge-base with data such as that generated by Fellows(Fellows, 1995), or by the generation of heating functions from exact solutions. The mode tuning software can also be improved by optimizing the cavity length and probe depth scaling parameters. The software platform, LabView is a very comprehensive tool and most importantly transparent to operating systems and thus, it is highly recommended as the best control software development tool available and must be kept 176 177 8. 2 Technology Advancements For further advancement, the variable frequency technology in addition to high power processing, dual coupling, scale-up and tooling issues must be actively pursued. These areas would provide speed and versatility to the technology which would enhance the application potential of it 8. 3 Variable Frequency Results from the work discussed in this dissertation have indicated that long mode switching times can be an unfavorable factor in heating certain samples. Hence, to further enhance the application flexibility of the single-mode resonant technology it is recommended that variable frequency processing is studied and automated. In the variable frequency technology, the frequency of the microwave power source is manipulated to control the electromagnetic modes rather than the cavity volume as is done in the mechanical tuning(see Chapter 3). Technically, electronic frequency manipulations can be done faster than mechanical cavity volume adjustments. Furthermore, with the variable frequency technology “moving parts” associated with the mechanical cavity adjustments are eliminated. Additionally, by the nature of the single-mode resonant cavity technology, a wider variety of modes can be excited in the . variable frequency system than in the fixed frequency system for cavities with similar dimensions. Compared with the fixed frequency system, the variable frequency system is more involved electronically, so it is recommended that such a system be automated. In the automation, the control objectives would be similar to the fixed frequency system but the tuning would be different The frequency would be manipulated electronically to minimize the reflected power, rather than the 2-dimensional tuning required in the fixed frequency system. 178 The following are recommended for the advancement of a variable frequency system: 1. Obtain a variable frequency microwave power source and adapt to existing hardware. This source should be purchased as a unit that includes all circuit components to avoid additional components for the wide frequency band. Note that the components of the fixed frequency are typically rated for 2.4 GHz and 100 Watts of power. It is important to note that as the frequency increases the free-space wavelength decreases which enforces more strict operating requirements on the system. 2 Fundamental work to characterize the system(empty and loaded) 3 Automation of the system 4. Process control logic development and implementation 5 Application to the processing of polymers and composites With the development of the variable frequency method the application potential of this technology would be significantly enhanced. Faster processing and more electromagnetic modes will be available for a similar cavity dimension which would improve the ability to process complex shaped parts and also achieve high-speed processing capabilities. Additionally, there will be the capability for dielectric analysis at various frequency ranges. 8.4 High power source To be able to process real parts at true high-speeds a high power source is highly recommended. The range of power required should be considered within the limitations of the processing hardware, sample properties and control capabilities. For the 3 inch square by 1 inch thick flat panel graphite / epoxy composites, preliminary heat transfer calculations show that a 200 W source could achieve heating rates of 60 °C /min. and a 1 KW source, 179 340°C / min. . Even with the electronic frequency switching system, such high heating rates as shown for the 1 KW source could pose tremendous control challenges. As such a 1 KW source may not be recommended for this sample. On the other hand, for a pultrusion process, preliminary technical, economic feasibility studies showed that a 1 KW source was required to achieve desirable throughput for processing 1 inch diameter polyester glass rods. Clearly the processing issues in these two processes would vary since one is a batch process and the other is a continuous process. Nevertheless, total process evaluations as well as hardware limitations must be understood in order to specify and define the high power source required for hi gh-speed processing. 8. 5 Dual Coupling As processing sample geometry becomes complex, the issue of near-fields could have a significant effect on heating. Hence, vertical coupling should be explored such that the near-field interactions can be avoided. This technology would require minimal effort since similar modes were found to be excitable with both vertical and horizontal coupling. Most importantly, it would provide additional processing flexibility for achieving uniform heating. As a novel approach, it could also be used in pultrusion applications by exciting orthogonal modes(degenerate modes) which can be used to process complex shaped parts or to uniformly process planar parts. To achieve this goal, further research must be done to understand this technology. The issue of phase locking the two sources must be explored to see whether dual coupling can be achieved at the non-degenerate cavity lengths. Also, dual coupling with various input power levels must be explored, such as each probe with equal power levels and with one having a higher power level than the other. 180 8.6 Scale-up Scale-up is another area which must be researched. In the single-mode resonant cavity technology, as the cavity size is scaled up the applied microwave frequency has to be scaled down in order to support similar electromagnetic modes. This means that for a 7 inch in diameter cavity which is used to process 3 inch square sample at 2.45 GHz, an 18 inch in diameter cavity would be required to process an 8 inch square sample at 0.915 GHz. Scale-up factors must be defined such that the 7 inch-diameter cavity results are used in larger sized cavities. These parameters would include microwave power frequency, sample placement, power density, cavity length, probe depth and tooling design. Currently, cylindrical shaped shaped cavities are used due to the ease of characterizing them, mathematically. However, as processing sample size increases and geometries become complex cavity geometry may be worth understanding. It may seem logical to process square shaped samples in square shaped cavities and discs in cylindrical cavities. 8. 7 Tooling Issues In the current state of technology, the part to be processed is placed in a specialized tooling that is placed inside the microwave cavity. These specialized tooling require microwave transparent materials such as polymers and ceramics which may cannot withstand processing pressures or handling in a manufacturing environment. The characterization results showed that, although the Teflon tooling material could be considered to be microwave transparent, its size could significantly perturb the electric fields inside the cavity. These perturbations could cause unpredictable field characteristics which could affect the heating of the part in an adverse manner. As such, the concept of tooling-free processing in the single-mode resonant cavity should be studied. l 8 1 8. 8 Potential Applications In general, economic feasibility, scale-up and tooling issues are just a few of the fundamental studies that must be done in order to advance this technology to commercialization. In commercialization there are several potential single-mode applications or those that do not require several modes to achieve uniform heating where this technology would be suitable. It is recommended that the following applications be considered for further progression of this technology. Pultrusion Integration of of the single mold resonant cavity as the main heating and shaping source (and not the pre-heater) in a pultrusion process for high throughput and enhanced properties. This may be done using the dual- coupling method for pultruding complex shaped parts, frequency switching for high-speed processing of multiple rods or mats, and fixed frequency for sheets or thin samples. Powder prepreg sintering Integration of the single-mode resonant system as the finishing step in the powder pre-pregging process. Post-curing The post-curing of parts where the bulk of the reaction has been completed and the part does not undergo rapid physical and chemical changes. An example would be RTM parts which are typically post-cured in thermal ovens and takes up to 24 hours. Potential benefit would be economics due to the minimization of the post-cure time. Dwmg Integrated as the drying step before an RTM process or the drying of polymers such as PET(poly ethylene terepthalate) for other applications. 182 Potential benefit would be speed and the enhancement of part properties, since microwave can be potentially used to dry polymers to lower moisture levels. Plasma support Ideal for plasma support processes since intensive process tuning is not required during processing. Potential benefit would be convenience of operating the process. Finally, these are only a few of the processing challenges that lay forth in this technology. A strong effort should be directed towards the elimination of the invasive temperature sensing methods to make this a true materials processing technology. However, this can only come about from a solid fundamental understanding of the technology. Hence, it is also very important that fundamental issues in process modeling, electromagnetic modeling, dielectric modeling be also pursued in parallel to the process developmental efforts in order to develop a su'ong foundation of the technology. APPENDICES APPENDIX A AUTOMATION -HARDWARE A.l Stepper Motors A.1.l Description The stepper motor is an 8-lead motor which was available as a motor and driver unit (Super Vexta UMDZ68M-El.5) and was designed specifically as a high performance positioning device. It is designed with class B insulation materials and is capable of withstanding temperatures at the motor coil of 130°C(255 F) with no reduction in motor life. It has a holding torque capability of 118 oz-inch and variable step angles of 0.90 and 0.450 which corresponds to 400 pulses per revolution and 800 pulse per revolution, respectively. A1.2 Stepper Motor Driver The driver for the stepper motors provide for switch selectable functions for step angle size, current cutback, automatic thermal cutout, pulse selection and direction selection, see Figure A-l. When SW-l and SW-2 are switched to the off position and SW-2 and SW—4 are switched to the on position, the motor would 183 184 W __ (43 + E IPulse G) + SW1 6') C CCW 1'1 ‘=- __ - 1.2 _= SW1 1.3 =- 2'1 11.; :- ® Striped——— 2.2 ' * o @5‘ Black ° —— Red :9 g z 1:} 6') 2 a, @ Orange _ Yellow £9— ~ AC115 0 ~:1 69 re F l IStriped J Motor Red Orange Yellow k J Figure A- l Stepper Motor and Drive Component require a high or low signal at the CW/CCW terminal to turn the motor in the clockwise or counterclockwise direction while pulses are sent to the pulse terminal. When SW-l and SW2 are switched on while SW-2 and SW-4 are switched off, the motor would require a pulse at the CW/CCW terminal to move in the clockwise direction and a pulse at the pulse terminal to move in the counter clockwise direction. The latter configuration is used in this work due to the convenience in programming. 185 The switch SW-S is used to select the step angle of the stepper motor to be either full step or half step, where a full step is 0.9 0 and a half step is 0.45 0 for every pulse. In this work both different step sizes are used for the different components depending of the travel length and speed. The switch SW-2 is used to select the option for automatic current cutback at standstill. When SW-2 is in the off position automatic current cutback is activated. This option allows for the reduction of power to a fraction of running power after about 200 milliseconds of inactivity. It also has a capability for T'I'L(Transducer Transducer Logic) digital logic IC(integrated circuit) interface for automatic motor pulse and direction selection. In a TTL type interface the voltage ranges from 0.0 V to 5.2 V, where high is defined as a voltage value being greater than +2.4 V and low to be less than +0.8 V. However, in the driver internal circuitry there is an optocoupler device which is used to drive the direction and pulse terminals which require 4-5 V for high and 0-0.5 V for low signals for TTL interface. Thus, a standard TTL wiring could not be used and a 5V external power supply(Tucker Power Supply) was necessary to drive these tenninals(see Figure A-2). r H: 4~SV N L: 0-0.5 V I 220:: I V I Optocou pier I Pulse or I Direction 1 E +25mAMax l L J Figure A- 2 Stepper Motor Driver TTL Circuit Movement of the stepper motor is accomplished by regulating the number and frequency of pulses sent to the driver's terminals. The motor rotates one step for 186 each pulse received at the pulse terminal and the direction of rotation is controlled by the signal applied to the driver's ”CW/CCW" terminal. Note that direction and pulse terminals are determined by the switch settings in the driver. A signal at (4-5 V) is said to be high while a signal at (0- 0.5V) is said to be low. A.1.3 Wiring Diagram The wire connections between the stepper motor, the power supply and the data acquisition board is as follows. The positive end of the pulse and CW/CCW terminals are connected to the 5V power supply and the negative end of these terminals are connected to the data acquisition terminal block. The negative end of the power supply is connected to the ground on the terminal block. These connections are very important in order for the proper operation of the stepper motors. Note that the 5V power supply from the data acquisition board was not adequate to drive these terminals because of the current requirements and the non- conventional high low voltage signals for the TTL output of the optocoupler (see Figure A-l). Figure A-2 shows wiring schematic for stepper motor drivers with the data acquisition interface board. 187 ( Digital Input! Output Terminals 1 @@@@® 9@@@%® paE+ @—O I I l u _ l l 6‘2: ' ‘* 43 . . l I CWE +92% . . . 69" — .. — _¢ : _. r— - — — J |@ e I Tucker Variable K PowerSupply j Figure A- 3 Wiring Schematic for Stepper Motor Drivers 188 A.2 Data Acquisition Interface A National Instrument data acquisition and control board NBMIOl6H was used. This board is a high performance multi-functional analog, digital, timing input/output board for the Macintosh computer. This board has a maximum sampling rate of 47 K samples per second and has an external circuit breaker which is attached to the side of the computer. It is a 12-bit successive approximation ADC(analog to digital converter) with 16 analog inputs, two 12-bit DACs with voltage outputs, 8 lines of TTL-compatible digital I/O, three l6-bit counter/timer channel for timing I/O, 2-(5 V) power supply, l-digital ground, l-analog input ground, and l-analog output ground(see Table A-2). This board supports three different input modes for the analog input channels which can be selected by jumpers W1 and W9. They are referenced single-ended (RSE) input, non-referenced single ended input (NRSE) input, and differential (DIFF) input. The referenced single-ended input configurations provides for 16 single ended channels, with the negative end of the of the input instrument referenced to analog ground. The differential ended input provides for eight channels with the negative end of the input instrument wire tied to the multiplexer output of channels 8 through 15. The non-referenced single-ended input configurations provides for 16 single ended channels, with the negative end of the input instrument tied to AISENSE and not connected to the ground. The RSE configuration is used for the analog input signals in this work. This configuration corresponds to jumper settings: A-B, C-D, G-H, and B-C. All jumper settings used in this work are summarized in Table A-1. The board also supports different input polarity and input ranges for the analog input signals. The two input polarities are unipolar and bipolar inputs. Unipolar input means the input voltage range is between 0 and the (voltage 189 reference- 1 LSB V), where the voltage reference is a positive reference voltage. Bipolar on the hand means that the input voltage range is between the negative of the voltage reference, and the (positive of the voltage reference - l LSB V). One LSB is the voltage increment corresponding to a least significant bit change in the digital code word. For unipolar output, 1 LSB = (Voltage reference/4096) and for bipolar lLSSB = (Voltage reference / 2048). The jumpers W3 and W4 can be used to select the different ranges. The options for ranges are 0 to 10 V, -5 to +5 V and -10 to +10 V. The voltage range of-lO to +10 V is used in this work which corresponds to the jumper setting of A-B and A-B. There is also software programmable gain settings which provides for more accurate voltage readings for different instruments. Note that the input range selection is based upon expected input range of incoming signals. A large input range typically sacrifices voltage resolution. The NBMIOl6H board supports gain settings of l,2,4,8 which correspond to the actual input range and precision as summarized in table 4-2.l. Note that the precision values are based upon the least significant bit (LSB) of the 12-bit ADC. This means that the voltage increment corresponds to a change of one in the ADC 12-bit count. The board also supports different polarity and ranges for the analog output channels. The jumpers for these settings are W4 and W-5 for either analog output channels 1 or 0. The direction of the digital ports can also be configured for either input or output by software. The factory settings are that port A is for input and port B is for output These ports support a maximum voltage rating of 5.5 V with respect to digital ground. For T'I'L logic it supports high at 2 V minimum and low at 0.8 V maximum. The current load at the high input voltage is 40 uA and -120 trA at low input voltage. It also supports output voltage of 2.4 Vmaximum for high and 0.5 V 190 maximum for low output logic, and 2.6 mA maximum output source current for high logic and sinks 24 mA for low output logic. ’ :evice I Data Acquisition Board Jumper Settings Input Mo e Range / Polarity jumper Setting IFmper Setting Analog Input (AD-RSE W1 A-B, C-D W3 A-B W9 G-H, B-C W4 A-B Table A- 1 Jumper Setting for Data Acquisiton Board APPENDIX B VERTICAL COUPLING B. 1 Introduction In the design of the mechanized cavity, as an added feature which was not required for automation, the mechanized cavity was equipped with novel axial and radial mounted coupling probes. As was previously mentioned typically, either the axial or radial mounted probe configuration is used and not both. Empty cavity characterization results showed that similar modes can be generated for either probe mount configurations except for at the cavity length corresponding to a degenerate mode. Using both probe mounted configurations the cavity was characterized for heating graphite / epoxy and vinyl / ester glass composite materials. B. 2 Experimental The applicator used in this study was the 7 inch diameter single mode resonant cavity, but with coupling probes mounted at the top and the side of the cavity, as shown in Figure B-l. The drive mechanisms for the coupling probes were designed for simultaneous or independent coupling. The vertical or t0p mounted probe mounted such that it can be adjusted independent of the top shorting plate. Thus, at a fixed cavity length the t0p mounted probe can be adjusted to various lengths. The dimensions and design of the t0p probe was similar to the side probe except for that it is longer. The top-mounted probe was 4" long x 0.375" in diameter and the side-coupling probe is 3" long x 0.375 " diameter and was located at ( 0.25 of the operating wavelength ) 1.2" from the base of the cavity. 191 192 I I ‘ op Probe ”ll/IIIIIIIIIJ Drive I - Short-plate W W Drive I a . hort-plate inger Stock Top Coupling Probe Sargle , = f "" l I Side Coup 'ng Probe Figure B- 1 Dual-Couplig Single mode Resonant Cavity Two microwave power sources were used with each one dedicated to each coupling probe. The sources were both magnetrons operating at a frequency of 2.45 GHz with a maximum power of 100W. The sample temperature was measured using fiber 193 optic temperature probes and thermal paper which changed color from white to blue at 85°C. Thermal paper was placed between the sample and the Teflon block with four fiber optic temperature probes attached to the top surface of the sample. Since the thermal paper activated at approximately 85°C, the sample was heated until the highest temperature measurement from the four temperature sites was 100°C. The samples were placed at 1.2” from the base of the cavity on a Teflon block, as shown in Figure B-2. Figure B-3 show temperature probe placements and sample placemt in the cavity. Loaded cavity modes were located by a swept frequency method. u- S\\‘ — l—Zl 1 Thermal Paper Figure B- 2 Sample placement with thermal paper 194 Fiber Direction 4 b 4 2 1 3 Figure B- 3 Temperature probe placement Two different heating experiments were done, one where only one probe was used to couple the microwaves into the cavity, and another where both probes were used to simultaneously couple energy into the cavity at a fixed cavity length. For the single . coupling experiments, when one coupling probe was active in coupling microwaves into the cavity the other was retracted from the cavity. Input and reflected power were measured from both probes at all times. In the dual coupling experiments each probe was tuned independent of the other at a fixed cavity length. 195 B. 3 Results and Discussions Similar empty cavity modes were located at similar cavity lengths for the top and side mounted probes. The heating profile for these modes were also similar for both probe configurations. The graphite epoxy composite also showed similar heating profile for both probe configurations except for at a cavity length of 19.7 cm, where edge heating was achieved by the top mounted probe and center heating for the side mounted probe. Figure B-4 and Figure B-5 show the therrnographs of these heating profiles. For a single mode resonant cavity only unique modes can be generated at a given cavity length, except for in certain regions of the cavity where two modes known as degenerate modes can exist at the same cavity length. In the seven inch cavity, there are two hybrid modes TM(l 1x) and TE(01x) which can exist at a cavity length of either 11.283 cm or 22.256 cm. Which of these two modes dominates during excitation is a function of where the coupling probe is mounted. The vertical coupling configuration should excite the TM-mode while the side mounted configuration should excite the TB- mode. The edge heating result from the top mounted probe could suggest a TE-mode excitation and a TM-mode for the side mounted probe which is contrary to what was expected. Figure B-9 shows the heating profile when both probes were used to simultaneously couple energy into the cavity. In this profile the temperature is uniform across the sample, indicated by the darkened areas on the thermal paper. This heating profile looks like the summation of the heating profiles from the side and top mounted coupling probes. This result is evidence that a hybrid mode could have been excited at this cavity length. 196 It was only possible to do dual coupling at the regions where hybrid modes were excited. At the other regions of the cavity, it was not possible since one probe acted as a source and the other as a sink. In other words, as one probe coupled energy into the cavity the other coupled it out of the cavity even when it was retracted from the cavity. The reflected power during dual coupling was less than 10% of the input power. Table B-1 lists all of the processing parameters for the dual coupling experiment The probe depth for the top mounted probe was always longer than that for the side mounted probe. Since the sample was placed at 1.2” from the bottom of the cavity it was relatively far enough away that near field effects would not be significant. However, the long metallic rod in the cavity could have a significant effect in the perturbation of the fields and make the heating pattern unpredictable. Both Probes 19.9 31.9 40.37 < 5 < 8 Table B- 1 Graphite I epoxy heating data g Figure B- 4 Top probe - graphite epoxy at CL=19.1 cm F \ X J Figure B- 5 Side probe - graphite epoxy at CL=19.7 cm 198 \ . .. 1 Figure B- 6 Both probes - graphite epoxy at CL=19.7 cm 8.4 Conclusions An interesting finding was that, only at the cavity length corresponding to the degenerate mode can simultaneous coupling from both axial and radial mounted coupling probes be used to achieve heating. At the other cavity lengths as one coupling probe coupled the microwave energy into the cavity the other coupled it out of the cavity and the sample was not heated. Theoretically the side mounted probe should preferentially excite TE-type modes while the axial mounted probe excites TM-type modes. This may suggest ' that at the degenerate mode, each probe preferentially excites standing waves that are ' orthogonal to one other and do not destructively interfere with one other. However, at the non-degenerate modes, both the axial and radial mounted probes excite similar modes that may be out of phase with one other and cause destructive interference; which was indicated by one coupling probe coupling the energy out of the cavity. This would be a hardware related problem that can be remedied by phase locking the microwave sources attached to the axial and radial mounted coupling probes. 199 The heating pattern that was observed using the degenerate mode, was uniformly distributed across the sample surface and resemble the superposition of the degenerate modes. What is notable about the simultaneous coupling method is that it provides for an additional methodology for achieving uniform heating in adjuvant to the mode-switching method using only the radial mounted coupling probe. Results of this work are discussed in appendix. The axial mounted probe is also significant in that it can be an alternative to the radial mounted probe in order to address the near-field interactions experienced from the radial mounted probe. Cavity characterization results from chapter 4 indicate that the near- frled interactions can alter field symmetry which can result in unpredictable field characteristics.The side coupling configuration is typically used in polymer processing, while vertical coupling is used in plasma generation. APPENDIX C KNOWLEDGE BASED SYSTEM INTERFACE C. 1 Introduction The lack of fundamental process models and the complexity of the domain in this technology, motivated the application of knowledge-based systems technique using a generic task approach for the overall automation. Initial implementation of the fabrication unit was the modelling and controlling of the processing device for mode tuning(Adegbite, 1991 ;Adegbite, 1992 ; Adegbite, 1992 ) using FM. The knowledge-based system technique was found to be effective for high level tasks such as planning than for fundamental control tasks(Adegbite, 1992 ). It was found to be slow and complicated in performing fundamental control tasks. Hence, a knowledge based system technique was applied in a collaborative research effort. The knowledge-based system planner was developed as a core dissertation work by Decker(Decker, 1995). which was interfaced with an automated system developed as a part of this dissertation, specifically for interfacing with the KBS system. C. 2 Goals of the Planner This planner is a part of a global objective set for this project is the development of ' an intelligent system for manufacturing composite materials from the design concept to fabrication of the final part. This will compose of two units with one dedicated to material design and the other to fabrication. The output of the desig n unit will be material properties and baseline process parameters to achieve specified application requirements. This output will be used as input to the fabrication unit to process the composite part The overall architecture in Figure 01 would be made up of the following components: 200 201 «Mm. Material and ; Properties Sensed Parameters Process 1 Ex tations C0 ns GlObal pec Planner Process States Process Process Profiles Conditions Process r States Reactive Monitor Characterization L Set -Points Controller Sensed Parameters Adjustments \ Process Sensed Parameters V Figure C- 1 Global Composite Manufacturing Architecture 202 1. An automated composite design unit with the capability to select composite material components and fabrication methods, given final part application(Kamel and Sticklen 1990; Kamel, Sticklen et al. 1990; Kamel, Sticklen et a1. 1992; Sticklen, Hawley et a1. 1992), 2. A mathematical modeling unit that would relate mass, momentum, and energy balances to explicitly describe the process(Sundaram, McDowell et a1. 1993) 3. A fabrication planning unit which would use process heuristics and process history to specify the most optimum processing protocol(Decker, Adegbite et a1. 1993), " 4 . A fabrication process control unit for the real-time operation of the fabrication process by sensing and controlling the fabrication process. This is the focus of the work discussed in this proposal(Adegbite, Hawley et al. 1991; Adegbite, Hawley et al. 1992; Adegbite, Hawley et al. 1992; Adegbite, Hawley et a1. 1993; Adegbite, Wei et a1. 1993). C3 Interface with KBS-planner In this work a control system was develop to enable the implemntaion of the planner. The planner was implementd in a computer languae called SmallTalk and process ' control interface was implemented in LabView. Details of the SmallTalk software can be found elsewhere(Decker, 1995), and details of the LabView section is presented. Output from the planner to the automated system were electromagnetic modes and power levels required to control the curing process to achieve uniform heating. Input to the planner from the automated system were the sensed process variables. Hence, the automated system had capabilities for acquiring data and carrying out control tasks that were requested by the planner, where the planner was the control decision making unit. 203 Several experiments were done to successfully verify this unit in the processing of graphite composite materials. The results of this work can be found elsewhere(Decker, 1995). The noteworthy element of this work is the interfacing capability between the knowledge-based system planner that was implemented in a software language called SmallTalk, and the automated system that was implemented in LabView. The successful demonstration of this system in curing experiments is a significant accomplishment in the application of a knowledge based-system planner in the real-time control of a complex process, and motivates new directions in process in automation. It provides the potential for capturing and representing experiential knowledge which typically resides with operators. This can be further advanced to knowledge retention and learning with the intended goals to eliminate the invasive sensing methods in the microwave processing technology. The LabView program for achieving this goal is called Integration Demo. APPENDIX D ELECTRIC FIELD PLOTS D. 1 Description Appendix D contains the Mathematica program routines that were used in the generation of the elecuic fields density plots. In this routine the spherical coordinates are converted to cartesian coordinates before plotting the data. The equations polotted are the magnitude of the elecuic field components in equations 3—12 and 3—15. For these calculations the roots of the Bessel’s function and their derivatives are input in files that are called on the first line of the program. Below is a a Mathematica program code for generating these plots. . TM-mode n21 v=3 . 8318/3 . S m11[x_,y_] :2 (8.3.91.3 [n.v Sqrt: [3‘2 + y‘2]] Sin [n ArcCoo [x/Sqrt [3*2+y*2] ] ] ) *2 Densityrloctm1tx,y] , (x, -3 . S. 3 . 5} . (Y: -3 . 5. 3 . S) . P10:Po1nto->90 , lush- ”bloc . Axon - >raloc , rrm- ”1130] TE—mode III-1 V81.8‘12/‘ n11[x_,y_] :- (1/2(Bouoldl(n-1).v Sqrtlx‘z + y“2]l - Boo-olJ[(n+1),v Sqrtlxez + yezll) Co. [I meo- [x/Sqrt [x“2+y“2] l l l ‘2 Donsity?10t[fl11[x.yl . (x. -3 . S. 3 . 5} . (Y. -3 . S, 3 . 5} , P10:Po:l.nt:o->90 Juan-”11:0, Axon-fl’alooer-u’alco] 204 APPENDIX E SOFTWARE DOCUMENTATION E. 1 Introduction The LabView program code for all of the main VI and the sub-VI are not included in the appendix because an acceptable print quality program could not be generated. This is due to the graphical nature of the program, which does not allow forrnating to specify margins or to generate a clear black and white image. Hence the progam subroutines are desribed. Each program is printed with a hierarchy which indicates the other subroutines or sub-VI’s that the program calls. There is also a front panel and a diagram included in each program. There are several subroutines used and only the ones developed to support this work are presented. These sub-VI’s have been previuosly described as part of the main program in Chapter 5. Again, a basic understanding of LabView would be required to fully comprehend the program. A title and description of each Vi is given. 205 206 E. 2 Compcure3-Curing program 13.2.1 Program Description This is the main LabView program for implementing the complete control-100p for curing. It shows in greater detail the different level of sequence structures. Parts of this program were shown in Chapter 5, however this is a more detailed representation of it. E. 3 Surface.vi - implementation of tuning program 13.3.1 Program Description This is the tuning program for implementing the 2-dimensional simplex method. It also uses several levels of sequence structure and calls several developed sub-VI’s as shown in the program hierarchy. The algorithm and details were presented in Chapter 5. The other sub-VI’s are presented in detail in the sections that follow. E. 4 Replace.vi-replacement of simplex triangle vertices E.4.l Program Description This a sub-VI that is only used in the Surface. Vi or tuning program. It is used to replace the vertices of the simplex triangle when new values are determined. It specifically replaces the cavity length and pobe depth values corresponding to the BNW vertices in the simplex triangle(see Chapter 5). E. 5 Vertex.vi -calculation of intial vertex in the simplex triangle E. 5.1 Program Description This sub-VI is also used specifically in the Surface.vi program and it is used to calcualte the intial vertices of the simplex triangle. These vertices are determined by calculating the cavity length and probe depths corresponding to a best point, next to best, worst point as explained in Chapter 5 in the tuning control software development. 207 E6 CL/ PD / Pwr -scaling of cavity length, probe depth and power values E 6.1 Program Description This sub-VI is used specifically for scaling the cavity length, probe depth and power sensed data. It converts the voltage readings to actual vaules. It is used in any 100p where data is acquired. E7 Move Read.vi - adjustment of cavity length & probe depth and power sensing E. 7.1 Program Description This sub Vi is also used in the surfacevi program for data acquisition and cavity length and probe depth adjustment. It contains the software for the stepper motor drivers. It was generally used to read the reflected power once new cavity length and probe depths were determined in the tuning program. E. 8 Limits.vi - tuning limits for a mode E. 8.1 Program Description This sub-VI is used to keep the tuning program from jumping out of modes. It keeps the tuning within the cavity length and probe depth within set limits for a desired . mode. E.9 Contract.vi - contraction of simplex triangle in the tuning program E. 9.1 Program Description This sub-VI is one ot the logic ateps in the simplex algorithm It is used to contract the triangle as described in Chapter 5 in the simplex routine. 208 E. 1 0 Pwrcontrl.vi - power controller used in the curing program E. 10.1 Program Description This is the main program where the PID controller is implemented. The subroutines for implementing the PID algorithm were available from LabView. Thus the program irnplemntation involved the constuction of the control logic using these algorithms. LIST OF WORKS CITED LISTS OF WORKS CITED Adegbite, V., M. Hawley, et a1. (1991). ‘ ° - ' ' .ASM/ESD 1991 Advanced Composites Conference and Exposition, Detroit Adegbite, V., J. Wei, Larry Fellow, M. Hawley, (1995).. ACS Cincinnatti Valerie Adegbite, Martin C. Hawley, Dave Decker, and Jon Sticklen, "Automation of Microwave Processing of Graphite/Epoxy Composite Materials Using an Expert Systems Technique", W, 162, 425 (1992). V. Adegbite, M. C. Hawley, A. Kama], M. Pegah, and J. Sticklen, "Monitoring and Control of Microwave Processing Utilizing Functional Modeling. Preliminary Results from the Automation of the Empty Cavity Mode Tuning", W, 2042 (1992). Valerie Adegbite, Martin C. Hawley, and John Sticklen, "A Control Strategy for the Microwave Processing of Composite Materials", WW ') mmr -m.__'o. rn'n W hin 11” Antsaklis, P. (1994). W. IEEE International Symposium on Intelligent Control, Asmussen, J. H. H. Lin, et a1. (1987). “Single-Mode or Controlled Multimode Microwave Cavity Applicators for Precision Material Processing.” Wm: 1477-1486. ' Asmussen, J ., R. Mallavarpu, et a1. (1974). IEEE, Asmussen, J. and D. Reinhard (1986). W: Asmussen, J. and J. Root (1984). Am 44: 396. Asmussen, J. and J. Root (1985). [1.5. Eaten: N9. 4 fill] 588 : Beveridge, G. S. G. and R. S. Schechter (1970). Wee. New York, McGraw-Hill. Binner, J. G. P., Ed. (1991). WWW. Cambridge, Abington Publishing. 209 210 Boas, A. H. (1963). “How Search Methods Locate Optima 1n Univariate Problems.” W112 70 85 88 Bolin, B., E. T. Degens, et al. (1979). The global biogeochemical carbon cycle. The W. New York. John Wiley. Box, M. J. (1965). “A New Method for Constrained Optimization and a Computation with other Methods.” W31 8: 42-52. Campey, I. G. and D. G. Nickols (1961). Simplex Minimization. ICI Limited- Central Instr. Lab. Carpenter, B. H. and H. C. Sweeney (1965). W 72(14): 117. Ciriscioli, P. R. and G. S. Springer (1989). “Dielectric Cure Monitoring- A critical Review.” W 25(3): 35-42. Ciriscioli, P. R. and G. S. Springer (1991). “An Expert System for Autoclave Curing of Composites ” MW 25: 1542- 1587. Collin. R. (1966). Emndanonsmmjmmfingineedng McGraW-Hill- Day, D. R. (1986). “Cure Control: Strategies for Use Of Dielectric Sensors.’ ’31s; IntcmatinnaLSAMEEfixmmsinm: 1095-1103. Dhulipala Ramakrishna, Susannah Travis, and Martin C. Hawley, "Microwave Processing of Glass FiberNinyl Ester-vinyl Toluene Composites", MRS Symp, firm" g2, 431 (1992). Fellows, L. (1992). Microwave Processing of Polyirnides. Michigan State University. Fellows L. A. , M. Lin, and M. C. Hawley, "Feasibility of Microwave Processing by Frequency Switching for Implementation of Mode-switching Techniques", Tenth Annual Polymer Processing Society Meeting, Akron, Ohio, April 3-7, 1994. Fellows . L. A, V. O. Adegbite, and M. C. Hawley, "Determination and Control of ' Microwave Processing Cycles", AIChE National Summer Meeting, Seattle, WA, August 15-18, 1993. Fellows, Larry , Susannah Travis, and Martin C. Hawley, "Preliminary Comparison of Microwave heating of Complex—shaped Composites Reinforced with Conductive and Nonconductive Fibers", W, 262, 415 (1992). Fellows, Larry , Richard Delgado, & Martin Hawley “Processing of Complex Shapes with Single-mode Resonant Frequency Microwave Applications”, 26th Intemational SAMPE Technical Conf., Oct. 17-20 1994, Atlanta, GA Gourdenne, A., 3rd International Conference on Composite Materials, Paris, 2 1514 (1980) 211 Gutowski, T. G., T. Morigaki, et a1. (1987). “The consiladtion of Laminate Composites.” MW 21 172- 138 Halpin, J, J. L. Kardos, et a1. (1983). “Processing Science: An Approach for Prepreg Composite Systems ” Wm 55: Harrington, R. F. (1961). IimeflannenieEleetremagnetieEields. New York, McGraw- Hill. Hart, P. (1982). “Directions for AI in the 808.” W (79): Hart, 3. and D. Kranbeuhl (1992). “Intelligent Sensor-Model Automated Control of PMR- 15 Autoclave Processing” Whom: 224-230. Hawley, M. C. J. Jow, et a1. (1987). “Microwave Processing and Diagnosis of Chemically Reacting Materials 1n a Single Mode Cavity Applicator.” MW W302): 1435-1443. Hinrichs, R., J. and J. Thuen M. (1984). “Control System For Processing Composite Materials.” W: Hinrichs, R. J. (1983). “Control of Composite Cure Processes.” ASIM_SIE 797: 29-37. Huack, H. S. (1969). “Design Considerations for Microwave Oven Cavities.” IEEE (Jan/Feb): Jow, J. (1988). Microwave Processing and Dielectric Diagnosis of Polymers and Composites Using A Single-Mode Resonant Cavity Technique. Michigan State University. Jullien, H. and H. Valot (1985). Eolxmer (26): 506. Kamel. A and J Sticklen (1990) AnAflificiaUntelfiaenctheiDesisnloolfoflmin Wm. Proceeding of the 5th Technical Conference for Composites, Kardos, J. L. H. P. Dudukovic, etal., Ed. (1983) W .1111. ' 2.2111.'.d.x'. 1’ 1 11” '11-. Emessing. Composite Materials. Karmazsin, E. and P. Satre (1985). 111mm 93: 305. Kranbeuhl, D, D. Eichinger, et a1. (1990). “On- Line In-situ Control of the Resin Transfer Molding Process’ ’MnmanonaLSAMEESxmmmm: 825- 834. Kranbeuhl, D. 13., P. Haverty, et a1. (1988). ' 1 .Society of Plastics Engineers 46th Technical Conference Proceedings, Kranbeuhl. D. E.. P. Haverty. et a1. (1987). W112]: 'v ' ' . 42nd Annual Conference, Composite Institute, The Society of Plastics Industry, Session 22—D, Kranbeuhl, D. E., M. Hoff, etal. (1988). “Insitu Measurement and Control of Processing Properties of Composite Resins 1n a Production Tool.” Wye Smmsim:1276-1284. 212 Kranbuehl, D., M. Hoff, et a1. (1988). IntemefieneleMEESmmsium 33: 1276. Kranbuehl, D., P. Kingsley, et a1. (1992). “Sensor-Model Prediction, Monitoring and In- situ Control of Liquid RTM Advanced Fiber Architecture Composite Processing. ” 31m IntemafinnaLSAMEESxmmsium: 907- 913 Krioger. B. (1994). WWW Mat. Res. Soc. Symp. Proc., LeClair, S. R., F. L. Abrams, et a1. (1989). “Qualitative Process Automation.” ALEdem 3(2): 125-136. Lee, C. W. (1987). “Composite Cure Process Control by Expert Systems.” SAMPE : Lee, I. W. and G. Springer (1984). “Microwave Curing of Composites.” lenmeLef W15 18: 386-409 Lee, W. I. andG. S. Springer(1984). “Interaction of Electromagnetic Radiation with Organic Matrix Composites} ’LeumeleLCeansneMetecieb 18. 357- 386. Lee, Y. F. and C. Y. C. Lee (1989). Proc. of the ACS Div. of Polym. Mater. Sci. and Eng., Lewis. D. (1992). W. Microwave Vulcanization; A lesson in Business and Technology, Lin, H.-H. (1989). Theoretical Formulation and Experimental Investigation of a Cylindrical Cavity Loaded with Lossy Dielectric Material. Michigan State University. McNeil D. M, M ..C Hawley, and M. DeMeuse, "Comparison of a Microwave and Thermally Cured Thennosetting Polycarbonate", W, 2621 (1992). Mallavarpu, R., M. C. Hawley, et a1. (1978). W PS-3: 55. Mannering, B. (1992). Electromagnetic Field Colutions for the Natural Modes of a Cylindrical Cavity Loaded With Lossy Materials. Michigan State Univerrsity. May, C. (1983). “The Chemical Characterization and Processing Of COmpoistes. ” Em: ' andAnnhedflhemrstnt 55(5). 811- 818. Montgomery. C. G. (1947). MW. New York. McGraw Hill. Morgan, 8., L and S. Deming N (1974). “Simplex Optimization of Analytical Chemical Methods. ” Anelxneelfihembuy, 46(9): 1170-1180. Mijovic, J. J Wijaya, Macromocolues, 23(15), 3677 (1990) Mijovic, J, J Wijaya, Polymer Composites, ll 184 (1990) Nelder, J. A. and R. Mead (1963). “A Simplex Method for Function Minimization.” Cnmnumloumal 7: 308-313. 213 Pardee, W. J. and B. Hayes-Roth (1987). Intelligent Real Time Control of Material Processing. Palo Alto Laboratory, Rockwell International. Pozar, D. M. (1990). Mmexefingmmfing. Addison-Wesley Publishing Company, Inc. Ramo, S. and J. Whinery (1953). EieldsendflemenMedemfiedie. New York, John Wiley and Sons, Inc. Richardson, T. (1987). W. New York, Industrial Press Inc. Risman, P. (1991). “Terminology and Notation of Microwave Power and Electromagnetic Energy ” WW 26(4): 243- 249 Roberts, R. W. (1987). ' ' ' . 32nd International SAMPE Symposium, Saliba, T., S. Saliba, et a1. (1992). “In-Situ Cure Monitoring of Advanced Composite Materials.” : Schwartz, M. M. (1984). W. New York, McGraw-Hill Book Company. ServaiS. R. A.. C. W. Lee. et al. (1986). WW3 31st International SAMPE SYmposium, Shinskey, F. G. (1988). W. New York, McGraw-Hill. Springer, G. S. (1987). “Modelling the Cure Process of Composites.” W21 W: 776-787 Stephanopoulos, G. (1984). ' - fieetiee. Edgewood Cliffs, Prentice-Hall Inc. Strong. B A (1939) WWW ManufacmdanatedalaMethodaAnnlicatinns Dearbom, Society of Manufacturing Engineers. ' Qiu, Yunchang, Val Adegbite, and Martin Hawley “Microwave Processing Using Intelligent Frequency Switching”, accepted for presentation at IMPI 30th Annual Microwave Symposiums, 1995 Qiu, Yunchang, Val Adegbite and Martin Hawley , “Intelligent Composite Processing Using Variable Frequency Method: Preliminary Heating Resuts”,submitted to , Symposium of “Smart Processing of Materials” on the 27th International SAMPE Technical Conference, 1995. Qiu, Yunchang ,Val Adegbite and Martin Hawley “Intelligent Microwave Processing of Composites Using Variable Frequency”,submitted to 11th Annual Advanced Composites Conference & Exposition, 1995 2 14 Sudaram, A. (1994). Microwave Process Modeling. Michigan State University. Sticklen, J., Kamel, A., Hawley, M., & Adegbite, V. (1992). liehfleedng Cemmsite IEEE Expert. 7(2). 43 53 Trttman, B. (1988). “Advanced Processing of Composites.” MRSjnlledn : 21. Trivisano, A. J. Kenny, et a1. (1992). “Control and Optimization of Autoclave Processing of High Performance Composites.” MW 37: 1104-1116. Vogel, G. L., J. Jow, et a1. (1989). 4th Tech. Conf. Comp. Mat... Wei, J. (1991). “Microwave Processing of Epoxy Resins and Graphite Fibe/Epoxy Composite 1n a Cylindrical Tunable Cavity.” Wei, J., J. Jow, et a1. (1989). AIChE Annual Conference, SF, CA, Wei, J ianghua and Martin C. Hawley, "Modeling and Controlling During Microwave and Thermal Processing of Composites", Wine, 2&2, 439 (1992). Wei, Jianghua,Mark DeMeuse, and Martin C. Hawley, "Kinetics Modeling of Microwave and Thermal Cure for Epoxy Resin", AQSJneeling, SF, CA, April, (1992). Wei, Jianghua , Brook Thomas, and Martin C. Hawley, "Scale-up Study of the Microwave Heating of Polymers and Composites", W, Anaheim, CA, March, (1992). Whitehair, S., J. Asmussen, et a1. (1987). l, Propel, Eewer 3(136): Whitehair, S., J. Asmussen, et a1. (1984). Annfledfimbieslenels 55: 1014. Wu, H.-T. and B. Joseph (1990). “Knowledge Based Control of Autoclave Curing of Composite.” SAMPE lenrnel 26(6): 39-54. Wu, H.-T. and B. Joseph (1990). “Knowledge Based Control of Autoclave Curing of Composites.” MM 26(6): 39-54. ' Young, P. R., M. A. Druy, et a1. (1988). “In-Situ Composite Monitoring Using Infrared Transmitting Optical Fibers.” 20m Intemetional SAMPE Tthical Cenferenee : 11-15. Young, V., J (1952). Undersgnding Mierowaves. New York, John F. RIder Publisher, Inc. "11111111111111ES